June 18, 2025 • 64 mins
Today's history lesson is about the non-markup language platform engineers love to hate, YAML Ain't Markup Language (YAML). Ingy tells us all about how and why it started, how it evolved over time, and what's happening next with YS.
Note: sorry about the audio issues in this episode. We did our best to clean it up.
Links:
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:01):
Hello and welcome to Fork Around and Find Out, the podcast all about forking around andfinding outs.
Today on the show, we are bringing you all stuff about YAML.
And if you uh don't know, we are gonna have some load-bearing white spaces in thisconversation.
Try, Autumn, don't take your head.
(00:21):
It's okay.
It wasn't a good joke, but I tried.
uh So I am excited that today on the show, we have ingy.net's...
uh
a creator and maintainer of the YAML spec and really excited to have you hear how itstarted, what you've been doing with it, and kind of what the future is.
Thanks Justin.
(00:41):
Thanks Adam.
Yeah, so my name's Ingy and I created YAML with Clark and Orin Beck.
We started in 2001.
Those guys were working on simplifying XML, which was going to take over the world around,you know, 1999, 2000, 2001 era and beyond.
(01:03):
But yeah, and so they were part of this.
mailing list called SML and simplify XML and then they kind of got sick of it and Clarkwas a Python guy so he made up something that looked like Python and he called it yet
another markup language and YAML and he got yaml.org and interestingly I had I was workingin a place called ActiveState up in Canada in Vancouver and they were famous for putting
(01:32):
Perl on Windows back in the late 90s and
Yeah, so at ActiveState, they decided they were going to be the enterprise based of opensource software, know, back in late 90s.
so, yeah, so they got like some of the top guys like the guy, the Perl release manager atthe time was Gurusamy Sarati and uh then the co-author of the Big Fat Python book from
(01:57):
remember that one from way back when the O'Reilly book.
Yeah, so David Asher and a bunch of other
you know, really heavy open source guys and me.
I just like it.
I've done something cool in Perl and got on the front page of the Perl Journal orsomething.
I saw these guys at OzCon and they had the coolest shirts on like, I want to go move up toVancouver from Seattle and work with those guys.
(02:20):
Anyway.
how to get into a project.
Have a cool show.
shirts, because I feel like this is important to the story.
they just had good branding after state data.
was like, it just said, A.
S.
and it kind of had this, this is kind of white stripes.
Right.
Because it had a, was before the white stripes.
had a well, black, red and white was kind of the, the color theme and it just looked cool.
(02:46):
Anyway, so I moved up to Vancouver and worked there, but at that time I had made a
A Perl serialization thing that looks like Python and it was called data denture forindented data.
It was just a CPAN model.
You know, it's still out there.
says go use YAML instead now.
somebody saw this SML project and data denture and wrote an email to the two of us.
(03:10):
It's like one line saying, Hey, you guys should meet.
So Clark and Clark called me on the phone.
gave my phone number.
He called me, we talked for four hours and YAML was on its way.
just like worked on it every day for the next five years over email.
So.
I don't know if you knew that, why would you?
yeah, Yama was entirely specked over emails about 20, 30 emails a day for five years.
(03:34):
And it was,
the original goalie though was people hated writing the brackets of XML.
Well, that was their take on it.
they wanted to...
People were starting...
XML wasn't really made for data, if I may say that, because it was made for textualmarkup, right?
That's actually a markup language.
(03:55):
And so they're trying to serialize data with it.
And data, even then, Perl and Python had the same hash array, scalar, or single value datamodel.
and then other languages follow suit.
you know, but I caught the JSON data model.
so, yeah, so XML doesn't fit well into that at all.
(04:18):
um You can try to fit.
Where do you draw the line there of like a markup language versus strictly data?
Well, OK, so like, you know, if I say ah I love Justin Garrison's podcast and I want tobold love, you know, then I would mark up the word.
ah But like in, you know, in data and in Python data side, you you wouldn't really do itthat way, right?
(04:46):
You do.
Yeah, yeah.
So Mark, these Markdown another that's another markup language trademark, obviously in thename.
So yeah, after like about six months of working on YAML, I'm like, you guys know thisisn't a markup language, right?
And they're like, yeah, it's a serialization language, which means, you know, write yourdata out to disk and read it back in or send it across the wire to some other program,
(05:09):
read it back into memory.
So we're like, yeah, but we already got the domain name.
We already got the name.
then I think Orin came up with it.
And Orin thinks I came up with it, but I'm like, how about, so one of us said,
GAML ain't markup language, you know, it's just the kind of recursive backronym.
Yeah, so.
(05:29):
What was the comparison or possibly competition with JSON?
Well, that's kind of funny.
you know, I thought JSON came out in 2006.
That's when we all saw it.
That's when the world saw it kind of overnight.
And it was so easy to implement that the implementation is just like, it's almost likewhen Tom came out a few years later, you know, somebody put out a spec and made it pretty
(05:53):
simple.
And then everybody jumped on it.
like, I want to be the GitHub inventor as a author of his markup language, which was, Idon't know if you ever saw the first spec of that, but it was like,
just a big attack.
was just anti-YAML, right?
It was like literally said, and then it ended with, and if you don't believe me, youhaven't had enough whiskey or something like that.
(06:15):
It was just a silly spec at the time.
Thank you.
I was just delivered coffee, so I had to take a sip there.
Okay, so.
ah
But JSON was around 2006, you all started this 2001.
so right, we started in 2001.
(06:35):
And then we looked at it and we're like, oh, that's our flow form looks exactly like thisJSON thing.
And then we're like, look closer and we're like, oh, it is exactly the same except forthree really minor details.
between a key value pair, you need a space after a colon and YAML.
And in JSON, you don't.
(06:57):
But JSON always uses the curlies.
for mappings and um it always quotes its keys.
It only uses string keys.
So we're like, okay, if you're in a curly brace thing and you have double quoted key, thenthere's not a space isn't required after the coding.
So then we're JSON compatible.
(07:17):
Is it like a minor thing that we did to the spec?
There's three really minor things that we did to the spec.
we made, so at YAML 1.1 was out and we went to YAML 1.2.
One of the main points of that was just to make
us a superset of JSON or JSON a subset of us officially.
And so, yeah, but I look back in history, JSON was started in 2001.
(07:39):
So, you know, it's just like, I think uh we started to get traction within the Rubylanguage, which I really credit for uh getting us off the ground.
You know, they shipped it with core probably 2003 or fours, so uh way too early, but youknow, they were just, they were kind of new kids on the block at the time and it was
pre-rails.
(08:00):
era and yeah.
What were some of the surprises along the way?
Like you mentioned getting in Ruby is one thing.
I feel like the YAML explosion of Kubernetes has been just, it became the thing everyoneloved to hate at some point.
And it's just like, we've got to write YAML.
Well, it's actually, it's been around for a long time.
Like what are those surprises along the way after 24 years of maintaining this?
(08:25):
Well, Justin, we actually met at at KubeCon.
don't remember if you like remember our reaction is like, think it was JJ from IBM andsaid, Hey, Justin, here's the the YAML guy.
And you like raised your fist, like you're going to, you know, hit me in the head orsomething like that.
And I knew you were kidding, but yeah, I'm like that guy.
(08:47):
And so this was my first KubeCon was last fall and I'm expecting
you cause you get all kinds of hate online rates, like, not a whole lot of love, you know?
So it's just like, is going to be terrible, but you know, I'll go on it.
That's fine.
And, um, and I got some exciting new stuff to share, but I went out and like, you know,besides like you and a couple other people, like, you know, mock hating.
(09:13):
was, was all love.
was like, it was crazy.
I mean, like you've got the people and it's like,
You know, they're kind of excited to meet the YAML guy and get a selfie, some of them, butthey're like, you know, they, truly like, I'm like, tell me what you don't like.
It's like, I like it.
It's great.
You know, I trying to explore, um, you know, and there's like, got to be just almostrunning on 20 competitors to the YAML crown.
(09:38):
As it were somebody at KubeCon calling that I didn't come up with that, but, uh, you know,and that's fair.
I mean, it's like, there's always, yeah, almost just one way to do things, but, uh,
It seems to be growing popular.
So you asked about along the way, the first one was Ansible.
you know, um all of sudden there was a company whose whole product was front-ended byYAML.
(09:59):
That was the first time I'd seen that.
I'm like, that's pretty cool.
um And then, you know, YAML was already like catching on.
It's just, it's just kind of, it's, I don't even know what part of the exponential curveit's on.
And it's not slowing down.
It should be.
but it's not quite yet and there's zero money behind it which is you know it's a goodtestament but also
(10:27):
I mean, should there be money behind specs?
Like that seems like any spec might die if there's money.
Like I look at HCL from HashiCorp, right?
Like it's their spec that they invented for simplified JSON.
And the more and more I see that evolve over time, like I don't think that has thelifespan of YAML because of just the...
(10:49):
could say all kinds of things about it.
It's like, don't want to offend Hashimoto or anything.
But I find it hilarious.
Well, kind of riffing off of, you know, all JSON is VML.
And actually there's an attack probably around 2018 where some guys like, there's theseedge cases uh having to do a surrogate pairs or the C1 section of uh unprintable
(11:16):
characters.
And we're like, OK, we had to get together.
have a group of five of us now that are on the core maintenance team.
And we got together and we like, and it was a hacker news thread.
And so people kept finding little edge cases.
And we had to take on, think, five of them.
And we're like, yeah, it's only a subset as far as we know.
(11:38):
ah We could be proven wrong.
And people come every once in while.
And we just turned out on the right side of it is.
a subset, you know, but it doesn't mean I kind of say, you know, by anybody up here whocan find that the case where it's actually not.
(12:02):
One beer.
but back to HashiCorp.
like JSON, if you don't like Hashi, HCL, you can use JSON, which means you can use YAML,but that is never mentioned by HashiCorp.
I think it's just like, they're like, we don't, because YAML was well out by the time allthis Hashi stuff came up.
(12:30):
Hashi, tools are great.
But it's like, you you could say that you could use the animal.
It's like, this would be people's choice.
If ATL is your preferred, then uh I never try to make technologies that like say, don'tuse his or hers, use mine.
Right?
I just say, I put myself out there and use it if you want.
(12:50):
So, type of thing.
But, so we'll talk about why's, this is how I pronounce the animal script is YS.
So it's like a.
Two letter Ys would be pronounced Ys, right?
But you can put it on it.
So I gave a talk on Helm-Ys, is YAML script applied Helm charts at that talk.
(13:12):
And that went over really well.
the thing that when people saw a side-by-side comparison of Ys, charts done with Yscompared to charts done with Go templating, which by the way, when you have Go templating
in your YAML files, they're no longer valid YAML files, right?
Not until they're rendered, but wise embedded is like makes them half the size.
(13:37):
That's the same thing.
And is valid.
YAML.
So you could use YAML tools in it.
Anyway, I'm kind of jumping ahead of that, yeah.
that gave me a question.
So I was a maintainer for uh a Helm chart for a very long time, and I hated Go templating.
uh
team now.
I have been since November.
I meet with Farina and all the guys every week.
was a Microsoft hackathon that I showed up at, which it was probably before a KubeCon atsome point.
(14:01):
I went to the Microsoft offices and they're like, we want people to do helm.
And so I learned how to write a helm charts.
And then I PR'd it to the repo and then I became the maintainer for it.
And I maintained it for years.
And I was like, I don't use this software.
I just used it to learn how to write a helm charts.
So hopefully it works for y'all.
But it was a thing that I was building.
I was like, I don't like this go templating.
(14:21):
I don't like templating in general in always.
always kind of has edge cases, right?
Like whether it's Go or Python or whatever, right?
Like at some form, like Ginges is going to fall apart just as easily as Go templates.
It's an attack vector, certainly in some cases.
I took a close look at the Go templating a couple of years ago and I'm like, I can't sayit's as bad as I thought it might be.
(14:46):
I looked for ways that it could go horribly wrong and I'm like, that's not too bad.
But then you're saying so like, why is a YAML script is an alternative for that helm wiseis an alternative for doing Go templating.
Yes, currently it's a post render.
You could actually go to search for Helm.YS.
anyway, it's a post render such that you can replace a little bit.
(15:10):
It'll work with the Go templating, or you could replace all the Go templating.
If you replace all of it, then your template files are valid YAML.
it's just a, if you were to go to yamlscript.org slash Helm.YS, that page is just a sideby side comparison.
can see it in front of face.
um
But let's-
I'm working with the Helm 14 to make it our first class citizen of without the postrendering factor if possible.
(15:41):
I guess let's take a step back.
What is the point of YAML script?
Yeah, it's a great question.
It's kind of funny because people like, you know.
(16:01):
People ask me to describe the use case for Uniscript.
It's like, well, you have an hour.
It's just like, it's like, there's not an elevator pits or I haven't got the perfect oneyet, but.
okay.
There we go.
So this year I focused on, wise is going to be focused on helping DevOps.
(16:22):
So I mean, there's a wise is a complete programming language with a complete ecosystembecause it builds on top of another language.
and inherits all of the goodness of that language.
There's a lot of great things about it.
But where it really can hit a mark for people is to use it in conjunction with whateverDevOps tools you're using.
(16:46):
even with that, like so say you had GitHub actions or CircleCI or a couple of that I'veplayed with, even though they don't load YAML files with with Wwise, which by the way,
YAML script
ships as a wise CLI.
That's one way to use it.
And it also ships as a YAML replacement loader in every language, where every is 11 sofar.
(17:09):
But I have a list of 45 languages that I want to ship this to in the next two years.
And they're really simple.
It's FFI binding to shared libraries.
it's really easy to create the next language I just need.
Usually I try to find somebody else to do it.
I'm like, you want it on?
This language, actually somebody did that last month at KubeCon.
(17:30):
They're like, Hey, do you have crystal yet?
I'm like, no, you want to add it?
And they did.
I got a pull request three days later.
So I don't know you've heard of the crystal language.
Yeah.
So that was really cool.
Yeah.
Um, and it gets more people involved.
try to get people involved in the project that way.
But yeah, it's basically think of all the things that people try to bolt onto YAML, right?
I mean, what's, what are the major problems with the animals?
(17:53):
You have these 5,000 line YAML config files, right?
I mean, it's really great to start your project in YAML where you're using it in a smallsuper clean and then it just starts growing and growing.
And then you have these 5,000 lines.
What if you could just compose it like you wouldn't write a 5,000 line Python program,right?
You break it up into libraries and subcomponents and all that.
Why is you can do that easily, right?
(18:15):
Or you can use modules from, it's not all 100 % working yet, but I will have modulesworking from
Either written in Wwise, written in Clozor, written in Java, in...
Because those are base things written in Wasm.
That's something I've been working a lot on, on different Wasm technologies recently.
(18:38):
um Or we actually right now we have a system where you can write it in any language, butthe linking mechanism is just basically sockets.
So it uses local sockets to communicate.
it's more of an R piece.
um
RPC type of thing.
basically, as far as you're concerned, you just load the SQL light mights, all thathappened to be written in Python and that just works, right?
(19:03):
So from a wise perspective.
But then you want interpolation.
want to be able to take different parts of your data and repeat uh loops on them.
And one of the problems with that with YAML, I'm sure you've heard of anchors and aliases,right?
Are you familiar with that feature of YAML?
That's where you have to give a label to
(19:25):
a node and then you can reuse it again later in the document.
So people will do things in pure YAML.
I mean, there's all kinds of frameworks.
The one thing about YAML being not quite finished the way we want it to is it opened upthe doors for dozens of projects to actually make money by, you know, fixing YAML for
(19:47):
their own specific use case.
I've always wanted to fix it for everyone.
I've always wanted to do, we set up to
You know, even Clark and Orin set out to replace all the XML landscape.
So that includes, um, you know, DSL languages, transformation languages, query languages,all the stuff that XML, all the extras that XML came with.
(20:09):
Um, yeah.
And then by the time we got the 1.2 spec out in 2009, they were just moved on to otherstuff and I stuck around, but it was around in, uh, 2023, all the pieces came together
where I'm like, oh, this is it.
I can make YAML script.
because all the components are there.
It was a combination of this closure language, is C-L-O-J-U-R-E, but it's pronouncedcloser.
(20:33):
And it's just a, it's a Lisp that compiles to JDM bytecode, so jar files.
And I had never programmed Java before and I'm like, I'm not going to, you know, go downthis route, but then there's another thing called growlVMNativeImage that will take any
jar thing and compile it to
(20:53):
uh native binary or a native shared library.
And for 20 plus years, I've been looking for the perfect language to write really highlevel code instead of C and make shared libraries such that I could bind it anywhere.
And then this all came together and I'm like, and if you look at the YAML script sourcecode, mean, it's really tight functional code.
(21:14):
So, and it produces this whole thing, but then I could instantly, within a few months, Ihad it bound to six languages, you know, and
I could concentrate on binding it to more languages.
always, but it's just kind of go with wherever people want me to take it next.
But the use case is still focused in for like DevOps professionals in that sort ofecosystem where people might be doing stuff with Kubernetes, uh where there is a lot of
(21:43):
YAML, something else on the like CACD pipelines, where it's GitLab or something else whereit's like this is defined in YAML as well.
And that's those areas because there's so much of it.
Is that why you're trying to just, you're trying to attack that as a place that peopleshould be using YAML.
like any problems that people have with YAML day to day, YAML script can solve about, youI'd say 90 plus percent of those.
(22:11):
my standpoint.
Is there anything that people have ever done with YAML that you weren't expecting, like ause case or just something that was completely out of left field?
Yeah, I mean, and that I love it.
Now I want to do it.
Okay.
So let me give you the example.
Um, comments in the YAML spec comments and YAML from the beginning were thrown away.
(22:32):
They don't fit into the model.
They don't fit into the JSON model.
like, and you can put them anywhere, right?
In a YAML document.
it's like, we wanted to keep comments around, but like now there's probably 50, 60, 70YAML frameworks out there.
That's what I would call a YAML implementation.
(22:52):
different languages.
Now there's probably half a dozen ones that will do comment roundtripping, but they um alldo it differently, right?
So um at this point I'm like, yeah, we need to actually put out a of a side spec of thisis maybe the preferred way to do it.
So just thought, this is related.
(23:17):
So something huge in YAML just happened.
three or four weeks ago, um Gustavo Niemeyer wrote GoYAML.
I didn't know this, but it's by far the biggest YAML asset in the world.
It's the YAML module that's behind all of Kubernetes.
um Even the Kubernetes SIGS YAML actually just has all of that code embedded in it becauseGustavo, three years ago, stopped maintaining it.
(23:46):
I've talked to Gustavo now and the reason he stopped maintaining it is because he has a uhvery high level job that demands his full attention.
It wasn't that he wanted to abandon it, but he just had to.
There's no time left in his daily life to work on this anymore.
So, and he said, I can't give it away because I don't know if it'll fall into disrepair.
(24:11):
And at first that's like, well, why can't you give it away?
But you know, after...
Reaching out to him and he's like, well, I could consider giving it to the YAMLorganization, but it's still kind of have the same problem.
So I'm like, well, we'll just take it over and run with it, but not as a, know, you won'tgive it to us.
We'll just make a fork as it were a copy and say that we're carrying this forward.
(24:35):
And if you want to, could, um, it's a couple of weeks away from launching this, but, um,officially like.
This is the new way to use GoYaml.
But it's going to be a huge boon to Go programming and the Kubernetes thing to have all ofthis unforked.
like the Kubernetes SIGs fork is going to use us as it's um upstream because, you know, Ireached out to them and I'm like, why don't you join us?
(25:03):
So they got one of their top guys.
You know, I just reached out to anybody who's really involved in this and had a stake init and was like, yeah, join the team.
You know, if...
um
will never break a release will never break your stuff because you're on the team.
You'll make sure it never breaks that kind of thing.
So that's really exciting.
eh
sure why people are more ready to put developers behind open source projects becausehaving a seat at the table and knowing that you won't get broken and being able to
(25:33):
contribute is huge.
Yeah.
And I was trying to tie this back to comments.
Gustavo, when I did talk to him for, he was so busy, we got to talk for 15 minutes.
But he said that he had ad comment support to go YAML, which I don't know.
And I don't think a lot of people use, it's a, yeah.
So comments was something that people just did.
(25:55):
you know, I mean, we didn't know what YAML was going to be for back in the day.
And we, we made a little list of use cases and anti-use cases and
configuration was in the anti, we're like, people won't want this for config.
I I don't know why we thought that at the time and that's all people use it for right now.
Yeah.
I think we're leaning towards grocery lists.
(26:18):
I don't.
at the Dammel script homepage, the examples here on the homepage have a lot of colons thatremind me of Pearl.
I don't know that's a closure.
I don't know if that's uh something coming through, but is there a reason that this, it tome feels like it's different or it's a lot different than.
(26:40):
than what YAML typically was.
When I first envisioned, when someone told me, when you told me YAML script, I was like,okay, this is gonna be like YAML with some like embedded commands, almost like bash has
escape shells, like subshells or whatever.
I'm just gonna have some back ticks in my YAML that does something.
This feels different with multiple colons and my imports and data.
(27:03):
Is there a reason you're going?
You want it to be a lot different, probably because of, to lose some historical baggage?
Is there a reason you want it to look like certain languages were adopted by something?
Is there something that inspired you to do it this way?
Well, I know what you're talking about.
you'll see a double colon instead of a colon between a key and a value.
(27:25):
Well, here's the deal.
Every YAML file that you know of is valid YAML scripts, right?
And it just does exactly what it did.
YAML script will load that perfectly.
And then you can use whatever the tool does.
can convert just loads of this data and you can convert it to different formats orwhatever.
But it's a YAML loader at the end of the day.
(27:47):
bang, YS at the top.
Then you said, okay, this is still YAML, but like, some of these things might be variablesor function names rather than strings, some of the unquoted things.
So any mapping array or scalar, those are the three nodes that make up the YAML or a JSONdocument can either be code or data.
(28:18):
So, you know, if you were just writing a program in YAML script, would be almost, youwouldn't really use data mode unless you needed a clean piece of data somewhere.
know, but for the use case that we're talking about where we're helping DevOps stuff,they're starting with their YAML files, but then they want to split it off and then load
in a different part, or they want to call a merge function with a specific thing.
(28:43):
So double colon says switch from.
The key, is in data mode, the value here is going to be code.
And it's going to be a code expression.
And there's a lot of really clean ways to do code expressions.
But then for the next key, it'll switch back to data mode in my view.
So that's why you see, if you look at those lines where you're seeing double colons,you'll see the thing on the left is probably just a string.
(29:11):
But it's unquoted because that's the clean way to write YAML.
You can put quotes around it.
And the thing on the right is an expression.
Yeah.
And the idea here is to allow people to incrementally adopt, right?
Like where you say, oh, can take out this one time.
have this one variable that I use eight times in my YAML file.
(29:32):
It was like, use the name of Kubernetes cluster, whatever it is, right?
It's like, oh, I could just store that as a value at the top and then replace that value.
So if I need to rename it or reuse this file or load from somewhere else, you can do that.
easily without saying I have to rewrite everything in this new language.
Yeah, let me tell you a little story.
(29:52):
and this is really cool.
And I've done this with a number of CID systems.
So let's take GitHub actions.
So um a friend of mine had a really hairy set of workflows and I looked through them andI'm like, yeah, YAML script can help out here.
So what I did was um I first, and there's a number of ways you can do this, but like,let's say you put all of the YAML script files in one directory, but they're going to
(30:18):
generate
the original YAML that you had from clean sources, right?
So the first step, if I take on a brand new repo of doing this, is I would copy all of theYAML files, which are YAML script, I've already said that, into the new source place.
And then I will call YAML script on those to gen all of the files in the place wherethey're supposed to be.
(30:42):
And then one time I will do a, I have a little script that will make sure that they'resemantically equivalent.
Was there something about that conversion that made them not be semantically equivalent?
Even though they're syntactically correct, they could be slightly different somehow.
yep.
And then once I've done that, then cleaning up, I just make the sources look as beautiful.
(31:06):
I try different things.
I'm like, oh, he reuses this a bunch of times.
I'm to put this in another file and then include it.
He uses all this repeating things.
I can make a function by just calling that function 20 times instead of.
That big blob of data.
and, but every time I do this, I'll make a small change and then I'll run the, you know,make update or whatever, however, I, um, the command that I did that will convert all the
(31:29):
files.
And if there's any diff in the result files, then I've done it wrong, right?
I haven't cleanly refactored.
So it's just a great workflow.
And then I just, you know, you know, I'll just spend like a big whack up time doing this.
And then I, you know, I presented it to him he's like, this is amazing.
And then he took it even further and.
Really dope, Anna.
I'll share that in the comments.
(31:50):
I assume there's comments.
We can add a link in the show notes.
Yeah.
Yeah, So remind me to do that.
How does that compare to the workflow that someone might have with like uh Q lang, right?
Like I look at Q as a way to generate YAML and I can say, I can take the output.
(32:11):
In that case, you're starting from scratch, right?
Cause Q lang is a different, it's not necessarily taking in the valid YAML, but I knowthere's a lot of...
tools and options that people have been trying to do.
like, well, I want to write real code or I want to write Python or I want to do somethingmore powerful.
Where's that trade off of the YAML script lets you start from what you have and slowlykind of change it over time versus one of those tools where it's like, I can go right to
(32:35):
what I think the end goal is of let my engineers write code in whatever language they wantand just generate the YAML from that.
Well, I think that you answered the question that you were asking.
mean, you were saying if you want to decide to do that, you know, if you're already aPython team and you're on a brand new project and we're going to like, yeah, we're going
(32:59):
to write Python code that generates our config.
me.
That's
I don't know if your video froze.
That's weird.
Let me try something.
Oh, experiments.
Are you back?
I'm still here.
I'm just seeing like a.
(33:20):
you now but your video is really slow and blurry.
Yeah, your video went real blocky and then I lost all audio for a bit.
should we do?
What do you, I could leave and come back like that, but I don't know if that would messyou up.
Okay.
Yeah, yeah, yeah.
Sure.
We can try that.
So we all kill it or just me.
I can, I mean, can kill it too, let's see.
(33:42):
Let's see if that audio is good.
I hear you much better now.
Sweet, let's start, you can start that answer, that question again.
Yeah, I don't remember the question.
Uh, starting either with, you know, from where they're at with YAML script and saying, orwith YAML and saying, I want to go from, from YAML to YAML script and clean this up or
(34:03):
starting with something like a QLang or something else that's is a, is a all from scratchsort of rewrite that generates the same YAML, which can give you the same workflow that
you're describing, but in different contexts, right?
Where one is iterating in checking versus starting from scratch and checking.
Yeah, so Qlang is, I mean, it's awesome.
(34:23):
So I look at it and I look for stuff to steal, right?
I mean, that's how you make a good language.
You look at what's great and you see if that fits into your language.
um YQ is another tool that's really great and that uses YAML and does things that YAMLscript can't do yet.
(34:43):
um And YAML script does things that YQ can't do.
um
When I actually ask the LLMs, I'm like, tell me about Q.
it's like, here's the advantages and disadvantages.
And some of the disadvantages would be that people really are used to YAML and they haveto learn this new system.
(35:06):
But if you're starting as a Q shop, great.
If there was a tool that could turn all of your YAML files into Q, and maybe there is, Idon't know, great.
That's a free one.
ah Whoever runs Q.
I that idea for free.
But yeah, so I, cause I do the same thing.
I'm like, you know, somebody told me all the stuff that you're doing is great, but ifsomebody has YAML files that are 5,000 lines long and they work, they're not going to
(35:34):
switch to YAML script.
And I'm like, fair.
And they're like, unless you give them a tool that does it for them automatically.
And I'm like, yeah, good point.
So uh I think I want to make,
YAML is great, but it was extremely limited.
The whole point of YAML script is, or why is this that it...
(35:57):
it tries to make YAML as good as, better than Python, better than, thank you, um at leastfor certain things.
So, um and, you
I know.
I have no problem with doing that.
Like I said, I'm not trying to compete.
I'm just trying to make great stuff.
So oh I was talking to Amazon AWS the other day oh to a guy in there and I was askingabout his products.
(36:31):
And so one is the CDK, the cloud development kit.
And basically they're doing exactly what you said.
They have like
you know, half a dozen languages, um, that make YAML that, you know, your team mightalready be familiar with.
And I'm like, well, you should throw YAML script.
And as one of those languages, you don't have to throw the others out, you know, but I betYAML script would fit really nicely in there because like, like you said, if you're
(36:57):
starting and you already have YAML, you kind of have to, it's like converting anythingfrom phone system another is usually kind of a waste of time in my opinion.
It's like,
A lot of companies will switch from like say a language like Ruby to Java or Go andthey'll come up with reasons, you know, because they can hire more developers or that kind
(37:22):
of thing.
And maybe those are true, but it's, know, you have a working system, build the new stuffin Go or Java or whatever, and keep your old stuff working if you really company wise.
I'm always, you know,
I feel like my experience, my experience is new things come in from the new.
(37:48):
uh
new projects, the new whatever initiative that says like, hey, we want to do the same andwe no longer write, we don't like Ruby anymore.
We're going to do Rust because whatever reason, right?
Like people are interested in it and that helps getting that excitement in a developerthat says like, hey, I get to write something new.
I really want to work on this project.
(38:08):
I don't want to write the same Rails app that I did for the last 15 years or whatever itis, but.
At some point, there's a tipping moment when enough of the new stuff is written in the newlanguage and you kind of lose the ability to modify and change and maintain the old
languages, whether because those people have moved on, whether because they just got sokind of kind of old and people didn't know how it worked anymore.
(38:36):
That's just like, wow, we got to have, we have one Ruby developer now that has to maintain18 different projects.
Fair, fair, but like, you know, when you first did the decision, just like, we're notgoing to be a Ruby shop.
We're going to be a Haskell shop.
I don't know.
I just made that up.
you know, we're going to be a Rust shop.
(38:57):
Does that mean you have to go and replace every piece of Ruby code, you know, because it'snot Rust?
No, it doesn't.
You know, it's like, you know, I just, I think a really smart CEO or CTO.
builds off of what they have and moves to the new thing.
And then it's like, yeah, we're losing all our Ruby stuff.
It's like, okay, well, maybe we need to take our more important things and convert themover now.
(39:21):
you know, it should never be the goal of a company that's trying to survive and make moneyand keep their people employed to, you know, replace every line of code from the old to
the new.
think that's personally.
Well, I don't
Most companies don't decide they're going to be a new type of language.
(39:42):
They're going to kind of experiment with it.
And then at some point they realize they are mostly writing in that new language with theengineers.
So it kind of evolves over time, but I also think that in general, at least with bigtechnology companies or, large companies in general, there's always a state of flux where
people are.
There's, there's always some re re-platforming or initiative to modernize or move eitherdocumentation to
(40:09):
a new wiki, code to a new CI CD, apps to new languages, like that's always happening andthings don't ever stay still.
And that's not necessarily a bad thing, but there is just always this like constant busywork that happens of none of this is making us money.
This doesn't.
that way.
mean, all you said is valid, but I think there's also times when you get a new CTO and hehas to make a name for himself.
(40:36):
So he's like, we're going to switch all of the developments of this.
that's the kind of thing that I think is, you know, then, yep, exactly.
Thank you, Adam.
that big impact to earn their giant stock piles.
uh
seems like right now people, were big tech companies are almost pushing like organizationsor companies as a whole over to whatever they decide they want to maintain.
(41:04):
know, Amazon used to be really Java heavy and now they're pushing JavaScript really hardand rest, which I mean, rest is a good language for a lot of things, but it seems like
they've picked what they want to invest in and kind of keep it going.
So.
And usually that involves some sort of vilification of the old, right?
Where you're like, oh, you're doing Java still?
(41:25):
That sucks, right?
You should be right in rest now because it's safe typed and whatever, you know, like, likeall the memory safety benefits and things that they can claim is like, you have to get to
the new thing.
sometimes it's bullying through like uh documentation, which I don't think they're doingthat.
I think Amazon will always be a Java shop, but I think for the new infrastructure, all ofthe documentation, well, not all of it, but a large majority of it and a large majority of
(41:53):
the examples are being done in JavaScript.
So that's kind of the option.
You know what I mean?
Cause there's so many examples in it that it's almost pushing teams that don't even
normally use JavaScript to do it for infrastructure, just like for a long time they wereusing Ruby for infrastructure.
That always confused me.
That probably is.
(42:14):
What do you think is the biggest money making language in the world right now?
I have a possible answer.
I think languages make money in a weird way though.
Like a lot of times it doesn't make money, it saves you money on compute and optimization.
Well, and you mean which are you just talking about like generally like which apps arewritten the most that make the most money like
(42:37):
in the world making the most money?
What language?
Yeah, I think uh it begins with a C.
I, yes, I was going to, so I told you about the person I met at KubeCon who wrote thecrystal binding for JavaScript, right?
(43:01):
I mentioned that.
so, yeah, I found that later.
uh
that she's a rock star at Red Hat working on IBM mainframes and she pointed me at anarticle.
uh I can link to this article.
It's a quote from IBM and so you've got to wonder how they got the numbers.
But if they're close, it's amazing.
(43:22):
68 % of production workloads are um running on IBM mainframes and of the entire cost ofall production workloads that they account for 6%.
So you're like, what the heck?
How's that possible?
Well, COBOL is the language of government, big pharma, medical, and banks, And insurance.
(43:48):
So the biggest industries in the world that are automated um are running these old systemsthat still work and they have to keep them working and they don't break and they don't
have any exposure to the internet because they run on systems.
that were pre-internet so that they don't get attacked, that kind of thing.
So it's really quite interesting that like a, I found this out while I was at a, so, youknow, being around from 2001, I was really big in the Perl community.
(44:18):
And to a degree, I, you know, I still, that's my family, right?
So there's a core developers meeting and it's in Europe every year they have in thespring.
the Perl toolchain summit and I was over there and I was like, hey guys, know, Perl wasmade to run, I just counted today, 25 different architecture platforms.
Like we don't even have that many platforms today, you know, but it was made in 87.
(44:42):
So it was ported to almost everywhere.
And I was like, do we still have access to, you know, IBM 360 architecture, which is nowcalled 390?
Do we have those machines left?
Because we had test machines for all this stuff.
They're like, no, we lost them a few years ago.
And I'm like, I think I know somebody who can get us access to those.
(45:05):
But in the back of my mind, I'm like, that would be cool to test on those things again.
what if Perl became the darling language of, uh know, because the new COBOL mainframes I'mtalking about, they've now made them run Linux, Kubernetes, seriously, ah all these modern
things, they're getting working on IBM mainframes.
(45:27):
What if
But not a lot of languages, other languages, are targeting that.
So what if Perl became the darling language of the COBOL mainframes again?
I don't know.
I thought it was quite an interesting possible fairy tale.
So we'll see how goes.
I actually met with that working group this morning for the first time to talk about howwe get Perl working full on with IBM stuff again.
(45:57):
I think uh the IBM testing goes through Oregon State University, which has mainframes.
And I remember just seeing that they were at risk of being defunded.
Did they get the funding?
Okay.
I know.
told that in Seattle here, there's a big mainframe center in South Seattle.
(46:18):
then, you know, a guy on the call today, he said he has access to one of these oldsystems.
I'm like, where is it?
like, it's in Dallas somewhere.
And then, you know, I applied to this Red Hat, you know, you can sign up for it to getaccess.
And yeah, there's like different people running mainframes.
And some of these are brand new, like mainframes.
you know, people carrying, it's like a big laptop, you know.
(46:41):
Mainframes used to be like big room size things, but now it's just like plugs in into a.
lot of military and government run mainframes.
Yeah, yeah, of course.
It's like.
there's some.
I always wonder if that's like if Java will hold on the same way that Cobalt does or ifCobalt is just so big.
Plus isn't tape getting really cheap also?
(47:04):
That's funny, tape.
um Did you say tape?
Yeah, we used to ship out, and I used to do mainframe assembler in the 90s.
We used to ship our releases on tape, and we used to ship our updates in a mailing, apaper mailing, and we'd send out, yeah, just if you want to fix this bug, just change
(47:25):
these lines 46 through, you know, whatever.
We'd give line numbers and lines of code of like, you know, just.
go change these files because we shipped the source because like it didn't really matterif you stole the source you'd never have the the machines to run it on right you know it's
like it's such a different world anyway
reading an article about somebody who used to work on mainframes, they were a TPM and justtheir kind of outlook on how like AI is changing everything, but how like everything
(47:57):
changed from like uh mainframes and then, you know, like server servers and clouds and allthat good stuff.
Well, I mean, I guess they're the same thing, but you know what I mean?
Like the kind of revolution of like how we use things.
It was really interesting.
You mentioned briefly about possibly helping people convert their YAML to YAML script.
(48:19):
Is that something that you're looking into?
Is that something that you think is going to be, we can take this giant 800 line Istiothousands of line Istio YAML Kubernetes file and convert it into YAML script.
Yeah, mean, yes, I'm thinking about it, like, you know, in the last, like this, you know,in 2025, I'm like, already got my fingers into like, got into Wazom, said, to, you know,
(48:47):
since KubeCon London into MCP, agentic AI stuff.
So it's like, you know, if, I think it might be important to experiment with, can you haveYAML script?
MC, MPC service sets that like, um, you know, some of this could be LL driven.
(49:07):
It would be really interesting to see if that was a possibility.
But yes, of course, if there was like, basically I work on a need basis right now, right?
You know, it's like, if people would be like, yeah, I would use this, but, uh, it's aproblem converting that, you know, and go down that alley or like, you know, definitely
try to find people that would want to work on that.
(49:30):
there's, there's so much cool stuff to.
to do uh with Wyze.
mean, it's just like, and with all these new technologies.
Yeah, it's exciting.
I made it.
I'm like, I gotta write a list because I have scraps of paper everywhere down.
I wrote a list of these are the kind of biggest projects I'm interested in.
(49:51):
And I thought it would be maybe a dozen.
It was like 26 as of this morning.
Like, the one cool thing about Yellowscript is I got so focused on it, I spent like 18months, you 12 hours a day.
just like, and I loved it every minute of it.
um Took all of last year off.
um Yeah, now currently.
(50:14):
You've doing technology for a while now.
How did you kind of stay interested in technologies from the 90s?
I mean, it's been what, 30 years now?
have you been doing the same thing of just finding something interesting and scratching anitch or are you...
(50:35):
day, I mean, it's just like, well, I made two important decisions.
like, don't get a big mortgage and don't.
I broke one.
I got married for just a little bit, but like, don't have kids.
Don't have pets and well, Ashley, my partner, she's we have plants now, so but she keepsthem alive, not me.
(51:01):
So, yeah, you know, I mean, I like.
You know, like, what is that line if do stuff you love and you'll never work a day in yourlife?
It's just, yeah, it's just like, like, you just take inputs and you're like, I don't know,it's still fun to me.
I don't know why maybe I'm just kind of a simple person.
Has there been any patterns you've seen that have been different for what's been fun?
(51:28):
Oh, yeah.
I see more patterns that keep it in similarities.
It's like always, you're not going to be super excited about what you're doing five yearsago.
Or maybe it morphed into something, right?
You're like, yeah.
(51:49):
Because I feel like a lot of the ideas are based off of other ideas.
it's probably all stuff I learned when I sat next to this guy who was super into
DRY, don't repeat yourself, programming.
And it's just like, I got my mind and I'm like, okay, this isn't dry enough.
How do we make it dry?
(52:09):
It's just like that.
And then you just like riff off into these wild frameworks and solutions.
And yeah, I think I wrote the first on JavaScript templating language.
It's called Jemplet back in 2006.
There wasn't anything that, and the cool thing about it is it actually compiled to.
JavaScript so you could have these template files and then you compile them to JavaScriptand JavaScript as pre node, right?
(52:36):
It's just starting to catch on like doing clean JavaScript in the browser instead of justhorrid stuff, but yeah.
I don't know.
And then somebody took that idea and ported it to like Lua and called it Lemplet orsomething like that.
It's just like, yeah, I don't know.
It's a...
(52:57):
It's kind of a wild ride, I do feel like when I saw wise, like, this is it.
This encapsulates.
That's why I said it's really hard to describe all the things.
Here's some interesting things.
I just noticed one day I'm like, it'd be really easy to use the tool chain that compilesclosure into a native binary since YAML script compiles the closure and then evaluates it.
(53:25):
What if I just take any YAML script program that you haven't wrote and you just say, know,YS tech tech binary that file and it just creates a native binary for you, things like
that.
So, um but also what if you could write
It's kind of that shared library.
was like, know, YAML script is even higher level, I feel like, than Clojure.
(53:49):
Clojure I kind of think of as my assembly language.
Sorry, Clojure does, but that's how I see it.
Because Lisp is a really normal type of way to generate code in, right?
It's hard to mess up.
Just keep your parentheses balanced, and you don't have to worry about much.
um So it's great that way.
um But with YAML script, I try to make it a...
I just program it in a lot.
(54:10):
And if I don't like something, it doesn't last long.
And I know that you've only seen a couple of examples.
The Helmwise stuff are kind of focused in a certain way.
But if I were to show you some other code, it's like, actually, I have a YAML scriptprogram that will take a
Okay, so there's a site called exorcism.org.
(54:32):
Do know that one?
It's a free language learning site.
It's great.
It's um, it's like exercise is um, so it's exorcism not exorcism.
Um, and it's it's great.
It's free.
It's uh, volunteer driven.
It's been around for a long time.
They have 75 languages.
One of them is YandleScript.
So I took all of last summer.
It's a lot of work to get a language learning track, YandleScript onto exorcism.org.
(54:59):
And this is where we cut because I was like.
all right.
Sorry.
I have to cut that out of it.
But I think it's been going pretty good otherwise.
Okay.
So, yeah, so there's also something called Rosetta code.
Are you familiar with Rosetta code?
(55:21):
Okay.
I actually like have met Mike.
Well, something happened this morning.
ah Rosetta code went down.
It might be down right now if you check.
And I reached out to Mike and I'm like, hey, your site's down.
And he's like, yeah, this is what's going on.
I'm like, I think I can help you when I get done with this.
you know, for.
For can find out podcast I'm So that's what I'm going help him get that back up.
(55:46):
If it's not up and running right now.
it's a domain registrar thing anyway.
um
Yeah, so I wrote something in 2012 or something.
I scraped all of Rosetta code.
it was like, and I do it like every, whenever I think about it, you know, it's not aprompt or whatever, but I like, it's about 120,000 individual program files.
(56:12):
It's like, I think they're coming up on a thousand languages, 950 plus languages against1500 programming tests in a matrix in a media Wiki, really janky, but like also pretty
cool.
it's, um,
You know, there's a Rosetta code data project on GitHub where you can just clone theentire thing.
And, um, so I took, I wanted to make sure that my exercise on solutions to, they have like150 standard problems that you know, are exposed a different amount in different
(56:45):
languages.
Python has a ton and you had to have 20.
I have 60 or YAML script has 60.
So I wanted to take these programs and compare them to.
uh code written in other languages.
And so I wrote uh a comparison program in YAML script that would, you know, just basicallymake a GitHub gist.
(57:06):
And you could do a little collapsible sections in GitHub.
it's with, you know, 74 other languages comparison, and then it shows the number of linesof code and like, how do I compare against Fortran or Go or Haskell or some really odd uh
ones.
anyway.
So that was a lot of fun.
I just, I'm like, I don't have to be the shortest program.
(57:31):
It's not golf language, but I just want to make sure that I'm in the right realm of, thisa good working language?
Is this how I would expect to program?
And I think it's quite close to that.
I used to write a new YAML script program and I would.
come up with 20 things I wanted to change.
Let's just look little things and now I, know, every once in a while I'll find one or two.
(57:55):
So.
the next five years of YAML looks like?
If we go ahead, it's 2030.
Where do think we're gonna be?
Yeah, with, I don't know, it's such a wild time, you know?
Like I'll keep pushing, like I think YAML script is the future of YAML.
(58:17):
And I'm just gonna quickly explain why.
So, because here's the problem with YAML is it's a giant spec.
It's very, very detailed and very easy to get wrong.
And in fact, we have a YAML test suite and in 2021, not a single language passed the testsuite and it wasn't even.
(58:38):
It was just the start of a test suite.
So it's like all of these implementations, they work good enough in their own language,but they just have all these little edge cases or minor discrepancies.
So they don't really work good between each other.
And that's what we originally signed up for.
At the same time, we wanted to put out the next version of YAML 1.3.
And we have a bunch of ideas of what we would do and even ways to do that without breakingthe world.
(59:06):
The problem with that is you put out a new spec version and it's taken people like yearsto just make the ones that they had.
Nobody's ever going to implement your next version of YAML.
It's not something anybody wants to do.
But what if there was one YAML loader that worked the same in every language, shipped toevery major language that uses YAML.
(59:26):
And the day that 1.3 or whatever the next YAML version spec version is, it was ready togo.
And it was full featured, did comments like we were talking about, did
everything and if there's a bug in Python, it had the same bug in Ruby and that same bugfix made by a Haskell person would affect that same person's bug in Java, right?
(59:48):
They'd all be the same.
So...
Whereas I you know, maybe the spec wasn't the best way to make you know, maybe it's a spinah
I don't know, implementation.
don't know.
I say both, right?
So it's like, I'll make the spec, people can still implement it.
And I don't think YAML script will be the only implementation to use, but it will be theone that will evolve and be the most correct with YAML for all languages.
(01:00:17):
So I think that's really important.
Even without the scripting, there's another, there's, can turn off all the scriptingcapabilities of YAML script.
You know, obviously.
You don't want to expose some of the things that it can do, like make web calls, databasecalls, whatever, or access file systems or parts of file systems.
(01:00:41):
Yeah, that's all configurable by either by both the user and by the consumers.
Well, Engy, thank you so much for coming on and telling us all about the past, present,and future of YAML.
I think it's been uh fascinating not even realizing that most of my career has beenevolved around YAML in some way, because I was writing Ansible pretty early on.
(01:01:06):
Even SALTs for config management was YAML based.
And while there was always this Python or JSON sort of bits and...
uh
other pieces it was all a subset of yaml in some way so that's really fascinating.
It's really rad, I didn't know YAML powered so many things.
(01:01:26):
Yeah, I didn't either to be honest.
mean, like, like when I went to KubeCon and just saw the amount of YAML in the world, justkind of messed with me a bit.
It's a bit crazy.
Where can people find you online?
there a place that you want people to find you online?
(01:01:46):
Absolutely, I'm like always, I can use another friend.
Yannelscript.org slash INGYDOTNET.
We can put that down in the links.
This is the page of all the ways to contact me.
So, it doesn't have my address, but if you ask me, come over and we'll do a barbecue orsomething.
(01:02:11):
Perfect, we'll have it in the show notes.
Thank you so much.
Yeah.
Autumn.
It's been awesome.
And you guys do a really good job of making a podcast flow really well.
really enjoyed doing this with you guys.
Thanks for coming on the show.
Yeah, we're just nerds that are interested and so it's fun.
(01:02:32):
It makes it really easy when there's an interesting topic.
When am I gonna, let's see, when am gonna see you again?
Are you going to, you guys got all the coupons?
Are North America's at least?
only wishes.
ah She doesn't tend to go to the cube cons.
hard and then all my friends have fun without me and then Justin sends me pictures.
It's so, it's, it's cruel.
cruel.
That's cruel, Justin.
(01:02:54):
I will be at KubeCon North America in Atlanta.
Yeah, that's in November.
So I will be there.
Sounds good.
Thank you everyone for listening and we will talk to you again soon.
And.
Hello and welcome to Fork Around and Find Out, the podcast all about forking around andfinding outs.
Today on the show, we are bringing you all stuff about YAML.
And if you uh don't know, we are gonna have some load-bearing white spaces in thisconversation.
Try, Autumn, don't take your head.
(00:21):
It's okay.
It wasn't a good joke, but I tried.
uh So I am excited that today on the show, we have ingy.net's...
uh
a creator and maintainer of the YAML spec and really excited to have you hear how itstarted, what you've been doing with it, and kind of what the future is.
Thanks Justin.
(00:41):
Thanks Adam.
Yeah, so my name's Ingy and I created YAML with Clark and Orin Beck.
We started in 2001.
Those guys were working on simplifying XML, which was going to take over the world around,you know, 1999, 2000, 2001 era and beyond.
(01:03):
But yeah, and so they were part of this.
mailing list called SML and simplify XML and then they kind of got sick of it and Clarkwas a Python guy so he made up something that looked like Python and he called it yet
another markup language and YAML and he got yaml.org and interestingly I had I was workingin a place called ActiveState up in Canada in Vancouver and they were famous for putting
(01:32):
Perl on Windows back in the late 90s and
Yeah, so at ActiveState, they decided they were going to be the enterprise based of opensource software, know, back in late 90s.
so, yeah, so they got like some of the top guys like the guy, the Perl release manager atthe time was Gurusamy Sarati and uh then the co-author of the Big Fat Python book from
(01:57):
remember that one from way back when the O'Reilly book.
Yeah, so David Asher and a bunch of other
you know, really heavy open source guys and me.
I just like it.
I've done something cool in Perl and got on the front page of the Perl Journal orsomething.
I saw these guys at OzCon and they had the coolest shirts on like, I want to go move up toVancouver from Seattle and work with those guys.
(02:20):
Anyway.
how to get into a project.
Have a cool show.
shirts, because I feel like this is important to the story.
they just had good branding after state data.
was like, it just said, A.
S.
and it kind of had this, this is kind of white stripes.
Right.
Because it had a, was before the white stripes.
had a well, black, red and white was kind of the, the color theme and it just looked cool.
(02:46):
Anyway, so I moved up to Vancouver and worked there, but at that time I had made a
A Perl serialization thing that looks like Python and it was called data denture forindented data.
It was just a CPAN model.
You know, it's still out there.
says go use YAML instead now.
somebody saw this SML project and data denture and wrote an email to the two of us.
(03:10):
It's like one line saying, Hey, you guys should meet.
So Clark and Clark called me on the phone.
gave my phone number.
He called me, we talked for four hours and YAML was on its way.
just like worked on it every day for the next five years over email.
So.
I don't know if you knew that, why would you?
yeah, Yama was entirely specked over emails about 20, 30 emails a day for five years.
(03:34):
And it was,
the original goalie though was people hated writing the brackets of XML.
Well, that was their take on it.
they wanted to...
People were starting...
XML wasn't really made for data, if I may say that, because it was made for textualmarkup, right?
That's actually a markup language.
(03:55):
And so they're trying to serialize data with it.
And data, even then, Perl and Python had the same hash array, scalar, or single value datamodel.
and then other languages follow suit.
you know, but I caught the JSON data model.
so, yeah, so XML doesn't fit well into that at all.
(04:18):
um You can try to fit.
Where do you draw the line there of like a markup language versus strictly data?
Well, OK, so like, you know, if I say ah I love Justin Garrison's podcast and I want tobold love, you know, then I would mark up the word.
ah But like in, you know, in data and in Python data side, you you wouldn't really do itthat way, right?
(04:46):
You do.
Yeah, yeah.
So Mark, these Markdown another that's another markup language trademark, obviously in thename.
So yeah, after like about six months of working on YAML, I'm like, you guys know thisisn't a markup language, right?
And they're like, yeah, it's a serialization language, which means, you know, write yourdata out to disk and read it back in or send it across the wire to some other program,
(05:09):
read it back into memory.
So we're like, yeah, but we already got the domain name.
We already got the name.
then I think Orin came up with it.
And Orin thinks I came up with it, but I'm like, how about, so one of us said,
GAML ain't markup language, you know, it's just the kind of recursive backronym.
Yeah, so.
(05:29):
What was the comparison or possibly competition with JSON?
Well, that's kind of funny.
you know, I thought JSON came out in 2006.
That's when we all saw it.
That's when the world saw it kind of overnight.
And it was so easy to implement that the implementation is just like, it's almost likewhen Tom came out a few years later, you know, somebody put out a spec and made it pretty
(05:53):
simple.
And then everybody jumped on it.
like, I want to be the GitHub inventor as a author of his markup language, which was, Idon't know if you ever saw the first spec of that, but it was like,
just a big attack.
was just anti-YAML, right?
It was like literally said, and then it ended with, and if you don't believe me, youhaven't had enough whiskey or something like that.
(06:15):
It was just a silly spec at the time.
Thank you.
I was just delivered coffee, so I had to take a sip there.
Okay, so.
ah
But JSON was around 2006, you all started this 2001.
so right, we started in 2001.
(06:35):
And then we looked at it and we're like, oh, that's our flow form looks exactly like thisJSON thing.
And then we're like, look closer and we're like, oh, it is exactly the same except forthree really minor details.
between a key value pair, you need a space after a colon and YAML.
And in JSON, you don't.
(06:57):
But JSON always uses the curlies.
for mappings and um it always quotes its keys.
It only uses string keys.
So we're like, okay, if you're in a curly brace thing and you have double quoted key, thenthere's not a space isn't required after the coding.
So then we're JSON compatible.
(07:17):
Is it like a minor thing that we did to the spec?
There's three really minor things that we did to the spec.
we made, so at YAML 1.1 was out and we went to YAML 1.2.
One of the main points of that was just to make
us a superset of JSON or JSON a subset of us officially.
And so, yeah, but I look back in history, JSON was started in 2001.
(07:39):
So, you know, it's just like, I think uh we started to get traction within the Rubylanguage, which I really credit for uh getting us off the ground.
You know, they shipped it with core probably 2003 or fours, so uh way too early, but youknow, they were just, they were kind of new kids on the block at the time and it was
pre-rails.
(08:00):
era and yeah.
What were some of the surprises along the way?
Like you mentioned getting in Ruby is one thing.
I feel like the YAML explosion of Kubernetes has been just, it became the thing everyoneloved to hate at some point.
And it's just like, we've got to write YAML.
Well, it's actually, it's been around for a long time.
Like what are those surprises along the way after 24 years of maintaining this?
(08:25):
Well, Justin, we actually met at at KubeCon.
don't remember if you like remember our reaction is like, think it was JJ from IBM andsaid, Hey, Justin, here's the the YAML guy.
And you like raised your fist, like you're going to, you know, hit me in the head orsomething like that.
And I knew you were kidding, but yeah, I'm like that guy.
(08:47):
And so this was my first KubeCon was last fall and I'm expecting
you cause you get all kinds of hate online rates, like, not a whole lot of love, you know?
So it's just like, is going to be terrible, but you know, I'll go on it.
That's fine.
And, um, and I got some exciting new stuff to share, but I went out and like, you know,besides like you and a couple other people, like, you know, mock hating.
(09:13):
was, was all love.
was like, it was crazy.
I mean, like you've got the people and it's like,
You know, they're kind of excited to meet the YAML guy and get a selfie, some of them, butthey're like, you know, they, truly like, I'm like, tell me what you don't like.
It's like, I like it.
It's great.
You know, I trying to explore, um, you know, and there's like, got to be just almostrunning on 20 competitors to the YAML crown.
(09:38):
As it were somebody at KubeCon calling that I didn't come up with that, but, uh, you know,and that's fair.
I mean, it's like, there's always, yeah, almost just one way to do things, but, uh,
It seems to be growing popular.
So you asked about along the way, the first one was Ansible.
you know, um all of sudden there was a company whose whole product was front-ended byYAML.
(09:59):
That was the first time I'd seen that.
I'm like, that's pretty cool.
um And then, you know, YAML was already like catching on.
It's just, it's just kind of, it's, I don't even know what part of the exponential curveit's on.
And it's not slowing down.
It should be.
but it's not quite yet and there's zero money behind it which is you know it's a goodtestament but also
(10:27):
I mean, should there be money behind specs?
Like that seems like any spec might die if there's money.
Like I look at HCL from HashiCorp, right?
Like it's their spec that they invented for simplified JSON.
And the more and more I see that evolve over time, like I don't think that has thelifespan of YAML because of just the...
(10:49):
could say all kinds of things about it.
It's like, don't want to offend Hashimoto or anything.
But I find it hilarious.
Well, kind of riffing off of, you know, all JSON is VML.
And actually there's an attack probably around 2018 where some guys like, there's theseedge cases uh having to do a surrogate pairs or the C1 section of uh unprintable
(11:16):
characters.
And we're like, OK, we had to get together.
have a group of five of us now that are on the core maintenance team.
And we got together and we like, and it was a hacker news thread.
And so people kept finding little edge cases.
And we had to take on, think, five of them.
And we're like, yeah, it's only a subset as far as we know.
(11:38):
ah We could be proven wrong.
And people come every once in while.
And we just turned out on the right side of it is.
a subset, you know, but it doesn't mean I kind of say, you know, by anybody up here whocan find that the case where it's actually not.
(12:02):
One beer.
but back to HashiCorp.
like JSON, if you don't like Hashi, HCL, you can use JSON, which means you can use YAML,but that is never mentioned by HashiCorp.
I think it's just like, they're like, we don't, because YAML was well out by the time allthis Hashi stuff came up.
(12:30):
Hashi, tools are great.
But it's like, you you could say that you could use the animal.
It's like, this would be people's choice.
If ATL is your preferred, then uh I never try to make technologies that like say, don'tuse his or hers, use mine.
Right?
I just say, I put myself out there and use it if you want.
(12:50):
So, type of thing.
But, so we'll talk about why's, this is how I pronounce the animal script is YS.
So it's like a.
Two letter Ys would be pronounced Ys, right?
But you can put it on it.
So I gave a talk on Helm-Ys, is YAML script applied Helm charts at that talk.
(13:12):
And that went over really well.
the thing that when people saw a side-by-side comparison of Ys, charts done with Yscompared to charts done with Go templating, which by the way, when you have Go templating
in your YAML files, they're no longer valid YAML files, right?
Not until they're rendered, but wise embedded is like makes them half the size.
(13:37):
That's the same thing.
And is valid.
YAML.
So you could use YAML tools in it.
Anyway, I'm kind of jumping ahead of that, yeah.
that gave me a question.
So I was a maintainer for uh a Helm chart for a very long time, and I hated Go templating.
uh
team now.
I have been since November.
I meet with Farina and all the guys every week.
was a Microsoft hackathon that I showed up at, which it was probably before a KubeCon atsome point.
(14:01):
I went to the Microsoft offices and they're like, we want people to do helm.
And so I learned how to write a helm charts.
And then I PR'd it to the repo and then I became the maintainer for it.
And I maintained it for years.
And I was like, I don't use this software.
I just used it to learn how to write a helm charts.
So hopefully it works for y'all.
But it was a thing that I was building.
I was like, I don't like this go templating.
(14:21):
I don't like templating in general in always.
always kind of has edge cases, right?
Like whether it's Go or Python or whatever, right?
Like at some form, like Ginges is going to fall apart just as easily as Go templates.
It's an attack vector, certainly in some cases.
I took a close look at the Go templating a couple of years ago and I'm like, I can't sayit's as bad as I thought it might be.
(14:46):
I looked for ways that it could go horribly wrong and I'm like, that's not too bad.
But then you're saying so like, why is a YAML script is an alternative for that helm wiseis an alternative for doing Go templating.
Yes, currently it's a post render.
You could actually go to search for Helm.YS.
anyway, it's a post render such that you can replace a little bit.
(15:10):
It'll work with the Go templating, or you could replace all the Go templating.
If you replace all of it, then your template files are valid YAML.
it's just a, if you were to go to yamlscript.org slash Helm.YS, that page is just a sideby side comparison.
can see it in front of face.
um
But let's-
I'm working with the Helm 14 to make it our first class citizen of without the postrendering factor if possible.
(15:41):
I guess let's take a step back.
What is the point of YAML script?
Yeah, it's a great question.
It's kind of funny because people like, you know.
(16:01):
People ask me to describe the use case for Uniscript.
It's like, well, you have an hour.
It's just like, it's like, there's not an elevator pits or I haven't got the perfect oneyet, but.
okay.
There we go.
So this year I focused on, wise is going to be focused on helping DevOps.
(16:22):
So I mean, there's a wise is a complete programming language with a complete ecosystembecause it builds on top of another language.
and inherits all of the goodness of that language.
There's a lot of great things about it.
But where it really can hit a mark for people is to use it in conjunction with whateverDevOps tools you're using.
(16:46):
even with that, like so say you had GitHub actions or CircleCI or a couple of that I'veplayed with, even though they don't load YAML files with with Wwise, which by the way,
YAML script
ships as a wise CLI.
That's one way to use it.
And it also ships as a YAML replacement loader in every language, where every is 11 sofar.
(17:09):
But I have a list of 45 languages that I want to ship this to in the next two years.
And they're really simple.
It's FFI binding to shared libraries.
it's really easy to create the next language I just need.
Usually I try to find somebody else to do it.
I'm like, you want it on?
This language, actually somebody did that last month at KubeCon.
(17:30):
They're like, Hey, do you have crystal yet?
I'm like, no, you want to add it?
And they did.
I got a pull request three days later.
So I don't know you've heard of the crystal language.
Yeah.
So that was really cool.
Yeah.
Um, and it gets more people involved.
try to get people involved in the project that way.
But yeah, it's basically think of all the things that people try to bolt onto YAML, right?
I mean, what's, what are the major problems with the animals?
(17:53):
You have these 5,000 line YAML config files, right?
I mean, it's really great to start your project in YAML where you're using it in a smallsuper clean and then it just starts growing and growing.
And then you have these 5,000 lines.
What if you could just compose it like you wouldn't write a 5,000 line Python program,right?
You break it up into libraries and subcomponents and all that.
Why is you can do that easily, right?
(18:15):
Or you can use modules from, it's not all 100 % working yet, but I will have modulesworking from
Either written in Wwise, written in Clozor, written in Java, in...
Because those are base things written in Wasm.
That's something I've been working a lot on, on different Wasm technologies recently.
(18:38):
um Or we actually right now we have a system where you can write it in any language, butthe linking mechanism is just basically sockets.
So it uses local sockets to communicate.
it's more of an R piece.
um
RPC type of thing.
basically, as far as you're concerned, you just load the SQL light mights, all thathappened to be written in Python and that just works, right?
(19:03):
So from a wise perspective.
But then you want interpolation.
want to be able to take different parts of your data and repeat uh loops on them.
And one of the problems with that with YAML, I'm sure you've heard of anchors and aliases,right?
Are you familiar with that feature of YAML?
That's where you have to give a label to
(19:25):
a node and then you can reuse it again later in the document.
So people will do things in pure YAML.
I mean, there's all kinds of frameworks.
The one thing about YAML being not quite finished the way we want it to is it opened upthe doors for dozens of projects to actually make money by, you know, fixing YAML for
(19:47):
their own specific use case.
I've always wanted to fix it for everyone.
I've always wanted to do, we set up to
You know, even Clark and Orin set out to replace all the XML landscape.
So that includes, um, you know, DSL languages, transformation languages, query languages,all the stuff that XML, all the extras that XML came with.
(20:09):
Um, yeah.
And then by the time we got the 1.2 spec out in 2009, they were just moved on to otherstuff and I stuck around, but it was around in, uh, 2023, all the pieces came together
where I'm like, oh, this is it.
I can make YAML script.
because all the components are there.
It was a combination of this closure language, is C-L-O-J-U-R-E, but it's pronouncedcloser.
(20:33):
And it's just a, it's a Lisp that compiles to JDM bytecode, so jar files.
And I had never programmed Java before and I'm like, I'm not going to, you know, go downthis route, but then there's another thing called growlVMNativeImage that will take any
jar thing and compile it to
(20:53):
uh native binary or a native shared library.
And for 20 plus years, I've been looking for the perfect language to write really highlevel code instead of C and make shared libraries such that I could bind it anywhere.
And then this all came together and I'm like, and if you look at the YAML script sourcecode, mean, it's really tight functional code.
(21:14):
So, and it produces this whole thing, but then I could instantly, within a few months, Ihad it bound to six languages, you know, and
I could concentrate on binding it to more languages.
always, but it's just kind of go with wherever people want me to take it next.
But the use case is still focused in for like DevOps professionals in that sort ofecosystem where people might be doing stuff with Kubernetes, uh where there is a lot of
(21:43):
YAML, something else on the like CACD pipelines, where it's GitLab or something else whereit's like this is defined in YAML as well.
And that's those areas because there's so much of it.
Is that why you're trying to just, you're trying to attack that as a place that peopleshould be using YAML.
like any problems that people have with YAML day to day, YAML script can solve about, youI'd say 90 plus percent of those.
(22:11):
my standpoint.
Is there anything that people have ever done with YAML that you weren't expecting, like ause case or just something that was completely out of left field?
Yeah, I mean, and that I love it.
Now I want to do it.
Okay.
So let me give you the example.
Um, comments in the YAML spec comments and YAML from the beginning were thrown away.
(22:32):
They don't fit into the model.
They don't fit into the JSON model.
like, and you can put them anywhere, right?
In a YAML document.
it's like, we wanted to keep comments around, but like now there's probably 50, 60, 70YAML frameworks out there.
That's what I would call a YAML implementation.
(22:52):
different languages.
Now there's probably half a dozen ones that will do comment roundtripping, but they um alldo it differently, right?
So um at this point I'm like, yeah, we need to actually put out a of a side spec of thisis maybe the preferred way to do it.
So just thought, this is related.
(23:17):
So something huge in YAML just happened.
three or four weeks ago, um Gustavo Niemeyer wrote GoYAML.
I didn't know this, but it's by far the biggest YAML asset in the world.
It's the YAML module that's behind all of Kubernetes.
um Even the Kubernetes SIGS YAML actually just has all of that code embedded in it becauseGustavo, three years ago, stopped maintaining it.
(23:46):
I've talked to Gustavo now and the reason he stopped maintaining it is because he has a uhvery high level job that demands his full attention.
It wasn't that he wanted to abandon it, but he just had to.
There's no time left in his daily life to work on this anymore.
So, and he said, I can't give it away because I don't know if it'll fall into disrepair.
(24:11):
And at first that's like, well, why can't you give it away?
But you know, after...
Reaching out to him and he's like, well, I could consider giving it to the YAMLorganization, but it's still kind of have the same problem.
So I'm like, well, we'll just take it over and run with it, but not as a, know, you won'tgive it to us.
We'll just make a fork as it were a copy and say that we're carrying this forward.
(24:35):
And if you want to, could, um, it's a couple of weeks away from launching this, but, um,officially like.
This is the new way to use GoYaml.
But it's going to be a huge boon to Go programming and the Kubernetes thing to have all ofthis unforked.
like the Kubernetes SIGs fork is going to use us as it's um upstream because, you know, Ireached out to them and I'm like, why don't you join us?
(25:03):
So they got one of their top guys.
You know, I just reached out to anybody who's really involved in this and had a stake init and was like, yeah, join the team.
You know, if...
um
will never break a release will never break your stuff because you're on the team.
You'll make sure it never breaks that kind of thing.
So that's really exciting.
eh
sure why people are more ready to put developers behind open source projects becausehaving a seat at the table and knowing that you won't get broken and being able to
(25:33):
contribute is huge.
Yeah.
And I was trying to tie this back to comments.
Gustavo, when I did talk to him for, he was so busy, we got to talk for 15 minutes.
But he said that he had ad comment support to go YAML, which I don't know.
And I don't think a lot of people use, it's a, yeah.
So comments was something that people just did.
(25:55):
you know, I mean, we didn't know what YAML was going to be for back in the day.
And we, we made a little list of use cases and anti-use cases and
configuration was in the anti, we're like, people won't want this for config.
I I don't know why we thought that at the time and that's all people use it for right now.
Yeah.
I think we're leaning towards grocery lists.
(26:18):
I don't.
at the Dammel script homepage, the examples here on the homepage have a lot of colons thatremind me of Pearl.
I don't know that's a closure.
I don't know if that's uh something coming through, but is there a reason that this, it tome feels like it's different or it's a lot different than.
(26:40):
than what YAML typically was.
When I first envisioned, when someone told me, when you told me YAML script, I was like,okay, this is gonna be like YAML with some like embedded commands, almost like bash has
escape shells, like subshells or whatever.
I'm just gonna have some back ticks in my YAML that does something.
This feels different with multiple colons and my imports and data.
(27:03):
Is there a reason you're going?
You want it to be a lot different, probably because of, to lose some historical baggage?
Is there a reason you want it to look like certain languages were adopted by something?
Is there something that inspired you to do it this way?
Well, I know what you're talking about.
you'll see a double colon instead of a colon between a key and a value.
(27:25):
Well, here's the deal.
Every YAML file that you know of is valid YAML scripts, right?
And it just does exactly what it did.
YAML script will load that perfectly.
And then you can use whatever the tool does.
can convert just loads of this data and you can convert it to different formats orwhatever.
But it's a YAML loader at the end of the day.
(27:47):
bang, YS at the top.
Then you said, okay, this is still YAML, but like, some of these things might be variablesor function names rather than strings, some of the unquoted things.
So any mapping array or scalar, those are the three nodes that make up the YAML or a JSONdocument can either be code or data.
(28:18):
So, you know, if you were just writing a program in YAML script, would be almost, youwouldn't really use data mode unless you needed a clean piece of data somewhere.
know, but for the use case that we're talking about where we're helping DevOps stuff,they're starting with their YAML files, but then they want to split it off and then load
in a different part, or they want to call a merge function with a specific thing.
(28:43):
So double colon says switch from.
The key, is in data mode, the value here is going to be code.
And it's going to be a code expression.
And there's a lot of really clean ways to do code expressions.
But then for the next key, it'll switch back to data mode in my view.
So that's why you see, if you look at those lines where you're seeing double colons,you'll see the thing on the left is probably just a string.
(29:11):
But it's unquoted because that's the clean way to write YAML.
You can put quotes around it.
And the thing on the right is an expression.
Yeah.
And the idea here is to allow people to incrementally adopt, right?
Like where you say, oh, can take out this one time.
have this one variable that I use eight times in my YAML file.
(29:32):
It was like, use the name of Kubernetes cluster, whatever it is, right?
It's like, oh, I could just store that as a value at the top and then replace that value.
So if I need to rename it or reuse this file or load from somewhere else, you can do that.
easily without saying I have to rewrite everything in this new language.
Yeah, let me tell you a little story.
(29:52):
and this is really cool.
And I've done this with a number of CID systems.
So let's take GitHub actions.
So um a friend of mine had a really hairy set of workflows and I looked through them andI'm like, yeah, YAML script can help out here.
So what I did was um I first, and there's a number of ways you can do this, but like,let's say you put all of the YAML script files in one directory, but they're going to
(30:18):
generate
the original YAML that you had from clean sources, right?
So the first step, if I take on a brand new repo of doing this, is I would copy all of theYAML files, which are YAML script, I've already said that, into the new source place.
And then I will call YAML script on those to gen all of the files in the place wherethey're supposed to be.
(30:42):
And then one time I will do a, I have a little script that will make sure that they'resemantically equivalent.
Was there something about that conversion that made them not be semantically equivalent?
Even though they're syntactically correct, they could be slightly different somehow.
yep.
And then once I've done that, then cleaning up, I just make the sources look as beautiful.
(31:06):
I try different things.
I'm like, oh, he reuses this a bunch of times.
I'm to put this in another file and then include it.
He uses all this repeating things.
I can make a function by just calling that function 20 times instead of.
That big blob of data.
and, but every time I do this, I'll make a small change and then I'll run the, you know,make update or whatever, however, I, um, the command that I did that will convert all the
(31:29):
files.
And if there's any diff in the result files, then I've done it wrong, right?
I haven't cleanly refactored.
So it's just a great workflow.
And then I just, you know, you know, I'll just spend like a big whack up time doing this.
And then I, you know, I presented it to him he's like, this is amazing.
And then he took it even further and.
Really dope, Anna.
I'll share that in the comments.
(31:50):
I assume there's comments.
We can add a link in the show notes.
Yeah.
Yeah, So remind me to do that.
How does that compare to the workflow that someone might have with like uh Q lang, right?
Like I look at Q as a way to generate YAML and I can say, I can take the output.
(32:11):
In that case, you're starting from scratch, right?
Cause Q lang is a different, it's not necessarily taking in the valid YAML, but I knowthere's a lot of...
tools and options that people have been trying to do.
like, well, I want to write real code or I want to write Python or I want to do somethingmore powerful.
Where's that trade off of the YAML script lets you start from what you have and slowlykind of change it over time versus one of those tools where it's like, I can go right to
(32:35):
what I think the end goal is of let my engineers write code in whatever language they wantand just generate the YAML from that.
Well, I think that you answered the question that you were asking.
mean, you were saying if you want to decide to do that, you know, if you're already aPython team and you're on a brand new project and we're going to like, yeah, we're going
(32:59):
to write Python code that generates our config.
me.
That's
I don't know if your video froze.
That's weird.
Let me try something.
Oh, experiments.
Are you back?
I'm still here.
I'm just seeing like a.
(33:20):
you now but your video is really slow and blurry.
Yeah, your video went real blocky and then I lost all audio for a bit.
should we do?
What do you, I could leave and come back like that, but I don't know if that would messyou up.
Okay.
Yeah, yeah, yeah.
Sure.
We can try that.
So we all kill it or just me.
I can, I mean, can kill it too, let's see.
(33:42):
Let's see if that audio is good.
I hear you much better now.
Sweet, let's start, you can start that answer, that question again.
Yeah, I don't remember the question.
Uh, starting either with, you know, from where they're at with YAML script and saying, orwith YAML and saying, I want to go from, from YAML to YAML script and clean this up or
(34:03):
starting with something like a QLang or something else that's is a, is a all from scratchsort of rewrite that generates the same YAML, which can give you the same workflow that
you're describing, but in different contexts, right?
Where one is iterating in checking versus starting from scratch and checking.
Yeah, so Qlang is, I mean, it's awesome.
(34:23):
So I look at it and I look for stuff to steal, right?
I mean, that's how you make a good language.
You look at what's great and you see if that fits into your language.
um YQ is another tool that's really great and that uses YAML and does things that YAMLscript can't do yet.
(34:43):
um And YAML script does things that YQ can't do.
um
When I actually ask the LLMs, I'm like, tell me about Q.
it's like, here's the advantages and disadvantages.
And some of the disadvantages would be that people really are used to YAML and they haveto learn this new system.
(35:06):
But if you're starting as a Q shop, great.
If there was a tool that could turn all of your YAML files into Q, and maybe there is, Idon't know, great.
That's a free one.
ah Whoever runs Q.
I that idea for free.
But yeah, so I, cause I do the same thing.
I'm like, you know, somebody told me all the stuff that you're doing is great, but ifsomebody has YAML files that are 5,000 lines long and they work, they're not going to
(35:34):
switch to YAML script.
And I'm like, fair.
And they're like, unless you give them a tool that does it for them automatically.
And I'm like, yeah, good point.
So uh I think I want to make,
YAML is great, but it was extremely limited.
The whole point of YAML script is, or why is this that it...
(35:57):
it tries to make YAML as good as, better than Python, better than, thank you, um at leastfor certain things.
So, um and, you
I know.
I have no problem with doing that.
Like I said, I'm not trying to compete.
I'm just trying to make great stuff.
So oh I was talking to Amazon AWS the other day oh to a guy in there and I was askingabout his products.
(36:31):
And so one is the CDK, the cloud development kit.
And basically they're doing exactly what you said.
They have like
you know, half a dozen languages, um, that make YAML that, you know, your team mightalready be familiar with.
And I'm like, well, you should throw YAML script.
And as one of those languages, you don't have to throw the others out, you know, but I betYAML script would fit really nicely in there because like, like you said, if you're
(36:57):
starting and you already have YAML, you kind of have to, it's like converting anythingfrom phone system another is usually kind of a waste of time in my opinion.
It's like,
A lot of companies will switch from like say a language like Ruby to Java or Go andthey'll come up with reasons, you know, because they can hire more developers or that kind
(37:22):
of thing.
And maybe those are true, but it's, know, you have a working system, build the new stuffin Go or Java or whatever, and keep your old stuff working if you really company wise.
I'm always, you know,
I feel like my experience, my experience is new things come in from the new.
(37:48):
uh
new projects, the new whatever initiative that says like, hey, we want to do the same andwe no longer write, we don't like Ruby anymore.
We're going to do Rust because whatever reason, right?
Like people are interested in it and that helps getting that excitement in a developerthat says like, hey, I get to write something new.
I really want to work on this project.
(38:08):
I don't want to write the same Rails app that I did for the last 15 years or whatever itis, but.
At some point, there's a tipping moment when enough of the new stuff is written in the newlanguage and you kind of lose the ability to modify and change and maintain the old
languages, whether because those people have moved on, whether because they just got sokind of kind of old and people didn't know how it worked anymore.
(38:36):
That's just like, wow, we got to have, we have one Ruby developer now that has to maintain18 different projects.
Fair, fair, but like, you know, when you first did the decision, just like, we're notgoing to be a Ruby shop.
We're going to be a Haskell shop.
I don't know.
I just made that up.
you know, we're going to be a Rust shop.
(38:57):
Does that mean you have to go and replace every piece of Ruby code, you know, because it'snot Rust?
No, it doesn't.
You know, it's like, you know, I just, I think a really smart CEO or CTO.
builds off of what they have and moves to the new thing.
And then it's like, yeah, we're losing all our Ruby stuff.
It's like, okay, well, maybe we need to take our more important things and convert themover now.
(39:21):
you know, it should never be the goal of a company that's trying to survive and make moneyand keep their people employed to, you know, replace every line of code from the old to
the new.
think that's personally.
Well, I don't
Most companies don't decide they're going to be a new type of language.
(39:42):
They're going to kind of experiment with it.
And then at some point they realize they are mostly writing in that new language with theengineers.
So it kind of evolves over time, but I also think that in general, at least with bigtechnology companies or, large companies in general, there's always a state of flux where
people are.
There's, there's always some re re-platforming or initiative to modernize or move eitherdocumentation to
(40:09):
a new wiki, code to a new CI CD, apps to new languages, like that's always happening andthings don't ever stay still.
And that's not necessarily a bad thing, but there is just always this like constant busywork that happens of none of this is making us money.
This doesn't.
that way.
mean, all you said is valid, but I think there's also times when you get a new CTO and hehas to make a name for himself.
(40:36):
So he's like, we're going to switch all of the developments of this.
that's the kind of thing that I think is, you know, then, yep, exactly.
Thank you, Adam.
that big impact to earn their giant stock piles.
uh
seems like right now people, were big tech companies are almost pushing like organizationsor companies as a whole over to whatever they decide they want to maintain.
(41:04):
know, Amazon used to be really Java heavy and now they're pushing JavaScript really hardand rest, which I mean, rest is a good language for a lot of things, but it seems like
they've picked what they want to invest in and kind of keep it going.
So.
And usually that involves some sort of vilification of the old, right?
Where you're like, oh, you're doing Java still?
(41:25):
That sucks, right?
You should be right in rest now because it's safe typed and whatever, you know, like, likeall the memory safety benefits and things that they can claim is like, you have to get to
the new thing.
sometimes it's bullying through like uh documentation, which I don't think they're doingthat.
I think Amazon will always be a Java shop, but I think for the new infrastructure, all ofthe documentation, well, not all of it, but a large majority of it and a large majority of
(41:53):
the examples are being done in JavaScript.
So that's kind of the option.
You know what I mean?
Cause there's so many examples in it that it's almost pushing teams that don't even
normally use JavaScript to do it for infrastructure, just like for a long time they wereusing Ruby for infrastructure.
That always confused me.
That probably is.
(42:14):
What do you think is the biggest money making language in the world right now?
I have a possible answer.
I think languages make money in a weird way though.
Like a lot of times it doesn't make money, it saves you money on compute and optimization.
Well, and you mean which are you just talking about like generally like which apps arewritten the most that make the most money like
(42:37):
in the world making the most money?
What language?
Yeah, I think uh it begins with a C.
I, yes, I was going to, so I told you about the person I met at KubeCon who wrote thecrystal binding for JavaScript, right?
(43:01):
I mentioned that.
so, yeah, I found that later.
uh
that she's a rock star at Red Hat working on IBM mainframes and she pointed me at anarticle.
uh I can link to this article.
It's a quote from IBM and so you've got to wonder how they got the numbers.
But if they're close, it's amazing.
(43:22):
68 % of production workloads are um running on IBM mainframes and of the entire cost ofall production workloads that they account for 6%.
So you're like, what the heck?
How's that possible?
Well, COBOL is the language of government, big pharma, medical, and banks, And insurance.
(43:48):
So the biggest industries in the world that are automated um are running these old systemsthat still work and they have to keep them working and they don't break and they don't
have any exposure to the internet because they run on systems.
that were pre-internet so that they don't get attacked, that kind of thing.
So it's really quite interesting that like a, I found this out while I was at a, so, youknow, being around from 2001, I was really big in the Perl community.
(44:18):
And to a degree, I, you know, I still, that's my family, right?
So there's a core developers meeting and it's in Europe every year they have in thespring.
the Perl toolchain summit and I was over there and I was like, hey guys, know, Perl wasmade to run, I just counted today, 25 different architecture platforms.
Like we don't even have that many platforms today, you know, but it was made in 87.
(44:42):
So it was ported to almost everywhere.
And I was like, do we still have access to, you know, IBM 360 architecture, which is nowcalled 390?
Do we have those machines left?
Because we had test machines for all this stuff.
They're like, no, we lost them a few years ago.
And I'm like, I think I know somebody who can get us access to those.
(45:05):
But in the back of my mind, I'm like, that would be cool to test on those things again.
what if Perl became the darling language of, uh know, because the new COBOL mainframes I'mtalking about, they've now made them run Linux, Kubernetes, seriously, ah all these modern
things, they're getting working on IBM mainframes.
(45:27):
What if
But not a lot of languages, other languages, are targeting that.
So what if Perl became the darling language of the COBOL mainframes again?
I don't know.
I thought it was quite an interesting possible fairy tale.
So we'll see how goes.
I actually met with that working group this morning for the first time to talk about howwe get Perl working full on with IBM stuff again.
(45:57):
I think uh the IBM testing goes through Oregon State University, which has mainframes.
And I remember just seeing that they were at risk of being defunded.
Did they get the funding?
Okay.
I know.
told that in Seattle here, there's a big mainframe center in South Seattle.
(46:18):
then, you know, a guy on the call today, he said he has access to one of these oldsystems.
I'm like, where is it?
like, it's in Dallas somewhere.
And then, you know, I applied to this Red Hat, you know, you can sign up for it to getaccess.
And yeah, there's like different people running mainframes.
And some of these are brand new, like mainframes.
you know, people carrying, it's like a big laptop, you know.
(46:41):
Mainframes used to be like big room size things, but now it's just like plugs in into a.
lot of military and government run mainframes.
Yeah, yeah, of course.
It's like.
there's some.
I always wonder if that's like if Java will hold on the same way that Cobalt does or ifCobalt is just so big.
Plus isn't tape getting really cheap also?
(47:04):
That's funny, tape.
um Did you say tape?
Yeah, we used to ship out, and I used to do mainframe assembler in the 90s.
We used to ship our releases on tape, and we used to ship our updates in a mailing, apaper mailing, and we'd send out, yeah, just if you want to fix this bug, just change
(47:25):
these lines 46 through, you know, whatever.
We'd give line numbers and lines of code of like, you know, just.
go change these files because we shipped the source because like it didn't really matterif you stole the source you'd never have the the machines to run it on right you know it's
like it's such a different world anyway
reading an article about somebody who used to work on mainframes, they were a TPM and justtheir kind of outlook on how like AI is changing everything, but how like everything
(47:57):
changed from like uh mainframes and then, you know, like server servers and clouds and allthat good stuff.
Well, I mean, I guess they're the same thing, but you know what I mean?
Like the kind of revolution of like how we use things.
It was really interesting.
You mentioned briefly about possibly helping people convert their YAML to YAML script.
(48:19):
Is that something that you're looking into?
Is that something that you think is going to be, we can take this giant 800 line Istiothousands of line Istio YAML Kubernetes file and convert it into YAML script.
Yeah, mean, yes, I'm thinking about it, like, you know, in the last, like this, you know,in 2025, I'm like, already got my fingers into like, got into Wazom, said, to, you know,
(48:47):
since KubeCon London into MCP, agentic AI stuff.
So it's like, you know, if, I think it might be important to experiment with, can you haveYAML script?
MC, MPC service sets that like, um, you know, some of this could be LL driven.
(49:07):
It would be really interesting to see if that was a possibility.
But yes, of course, if there was like, basically I work on a need basis right now, right?
You know, it's like, if people would be like, yeah, I would use this, but, uh, it's aproblem converting that, you know, and go down that alley or like, you know, definitely
try to find people that would want to work on that.
(49:30):
there's, there's so much cool stuff to.
to do uh with Wyze.
mean, it's just like, and with all these new technologies.
Yeah, it's exciting.
I made it.
I'm like, I gotta write a list because I have scraps of paper everywhere down.
I wrote a list of these are the kind of biggest projects I'm interested in.
(49:51):
And I thought it would be maybe a dozen.
It was like 26 as of this morning.
Like, the one cool thing about Yellowscript is I got so focused on it, I spent like 18months, you 12 hours a day.
just like, and I loved it every minute of it.
um Took all of last year off.
um Yeah, now currently.
(50:14):
You've doing technology for a while now.
How did you kind of stay interested in technologies from the 90s?
I mean, it's been what, 30 years now?
have you been doing the same thing of just finding something interesting and scratching anitch or are you...
(50:35):
day, I mean, it's just like, well, I made two important decisions.
like, don't get a big mortgage and don't.
I broke one.
I got married for just a little bit, but like, don't have kids.
Don't have pets and well, Ashley, my partner, she's we have plants now, so but she keepsthem alive, not me.
(51:01):
So, yeah, you know, I mean, I like.
You know, like, what is that line if do stuff you love and you'll never work a day in yourlife?
It's just, yeah, it's just like, like, you just take inputs and you're like, I don't know,it's still fun to me.
I don't know why maybe I'm just kind of a simple person.
Has there been any patterns you've seen that have been different for what's been fun?
(51:28):
Oh, yeah.
I see more patterns that keep it in similarities.
It's like always, you're not going to be super excited about what you're doing five yearsago.
Or maybe it morphed into something, right?
You're like, yeah.
(51:49):
Because I feel like a lot of the ideas are based off of other ideas.
it's probably all stuff I learned when I sat next to this guy who was super into
DRY, don't repeat yourself, programming.
And it's just like, I got my mind and I'm like, okay, this isn't dry enough.
How do we make it dry?
(52:09):
It's just like that.
And then you just like riff off into these wild frameworks and solutions.
And yeah, I think I wrote the first on JavaScript templating language.
It's called Jemplet back in 2006.
There wasn't anything that, and the cool thing about it is it actually compiled to.
JavaScript so you could have these template files and then you compile them to JavaScriptand JavaScript as pre node, right?
(52:36):
It's just starting to catch on like doing clean JavaScript in the browser instead of justhorrid stuff, but yeah.
I don't know.
And then somebody took that idea and ported it to like Lua and called it Lemplet orsomething like that.
It's just like, yeah, I don't know.
It's a...
(52:57):
It's kind of a wild ride, I do feel like when I saw wise, like, this is it.
This encapsulates.
That's why I said it's really hard to describe all the things.
Here's some interesting things.
I just noticed one day I'm like, it'd be really easy to use the tool chain that compilesclosure into a native binary since YAML script compiles the closure and then evaluates it.
(53:25):
What if I just take any YAML script program that you haven't wrote and you just say, know,YS tech tech binary that file and it just creates a native binary for you, things like
that.
So, um but also what if you could write
It's kind of that shared library.
was like, know, YAML script is even higher level, I feel like, than Clojure.
(53:49):
Clojure I kind of think of as my assembly language.
Sorry, Clojure does, but that's how I see it.
Because Lisp is a really normal type of way to generate code in, right?
It's hard to mess up.
Just keep your parentheses balanced, and you don't have to worry about much.
um So it's great that way.
um But with YAML script, I try to make it a...
I just program it in a lot.
(54:10):
And if I don't like something, it doesn't last long.
And I know that you've only seen a couple of examples.
The Helmwise stuff are kind of focused in a certain way.
But if I were to show you some other code, it's like, actually, I have a YAML scriptprogram that will take a
Okay, so there's a site called exorcism.org.
(54:32):
Do know that one?
It's a free language learning site.
It's great.
It's um, it's like exercise is um, so it's exorcism not exorcism.
Um, and it's it's great.
It's free.
It's uh, volunteer driven.
It's been around for a long time.
They have 75 languages.
One of them is YandleScript.
So I took all of last summer.
It's a lot of work to get a language learning track, YandleScript onto exorcism.org.
(54:59):
And this is where we cut because I was like.
all right.
Sorry.
I have to cut that out of it.
But I think it's been going pretty good otherwise.
Okay.
So, yeah, so there's also something called Rosetta code.
Are you familiar with Rosetta code?
(55:21):
Okay.
I actually like have met Mike.
Well, something happened this morning.
ah Rosetta code went down.
It might be down right now if you check.
And I reached out to Mike and I'm like, hey, your site's down.
And he's like, yeah, this is what's going on.
I'm like, I think I can help you when I get done with this.
you know, for.
For can find out podcast I'm So that's what I'm going help him get that back up.
(55:46):
If it's not up and running right now.
it's a domain registrar thing anyway.
um
Yeah, so I wrote something in 2012 or something.
I scraped all of Rosetta code.
it was like, and I do it like every, whenever I think about it, you know, it's not aprompt or whatever, but I like, it's about 120,000 individual program files.
(56:12):
It's like, I think they're coming up on a thousand languages, 950 plus languages against1500 programming tests in a matrix in a media Wiki, really janky, but like also pretty
cool.
it's, um,
You know, there's a Rosetta code data project on GitHub where you can just clone theentire thing.
And, um, so I took, I wanted to make sure that my exercise on solutions to, they have like150 standard problems that you know, are exposed a different amount in different
(56:45):
languages.
Python has a ton and you had to have 20.
I have 60 or YAML script has 60.
So I wanted to take these programs and compare them to.
uh code written in other languages.
And so I wrote uh a comparison program in YAML script that would, you know, just basicallymake a GitHub gist.
(57:06):
And you could do a little collapsible sections in GitHub.
it's with, you know, 74 other languages comparison, and then it shows the number of linesof code and like, how do I compare against Fortran or Go or Haskell or some really odd uh
ones.
anyway.
So that was a lot of fun.
I just, I'm like, I don't have to be the shortest program.
(57:31):
It's not golf language, but I just want to make sure that I'm in the right realm of, thisa good working language?
Is this how I would expect to program?
And I think it's quite close to that.
I used to write a new YAML script program and I would.
come up with 20 things I wanted to change.
Let's just look little things and now I, know, every once in a while I'll find one or two.
(57:55):
So.
the next five years of YAML looks like?
If we go ahead, it's 2030.
Where do think we're gonna be?
Yeah, with, I don't know, it's such a wild time, you know?
Like I'll keep pushing, like I think YAML script is the future of YAML.
(58:17):
And I'm just gonna quickly explain why.
So, because here's the problem with YAML is it's a giant spec.
It's very, very detailed and very easy to get wrong.
And in fact, we have a YAML test suite and in 2021, not a single language passed the testsuite and it wasn't even.
(58:38):
It was just the start of a test suite.
So it's like all of these implementations, they work good enough in their own language,but they just have all these little edge cases or minor discrepancies.
So they don't really work good between each other.
And that's what we originally signed up for.
At the same time, we wanted to put out the next version of YAML 1.3.
And we have a bunch of ideas of what we would do and even ways to do that without breakingthe world.
(59:06):
The problem with that is you put out a new spec version and it's taken people like yearsto just make the ones that they had.
Nobody's ever going to implement your next version of YAML.
It's not something anybody wants to do.
But what if there was one YAML loader that worked the same in every language, shipped toevery major language that uses YAML.
(59:26):
And the day that 1.3 or whatever the next YAML version spec version is, it was ready togo.
And it was full featured, did comments like we were talking about, did
everything and if there's a bug in Python, it had the same bug in Ruby and that same bugfix made by a Haskell person would affect that same person's bug in Java, right?
(59:48):
They'd all be the same.
So...
Whereas I you know, maybe the spec wasn't the best way to make you know, maybe it's a spinah
I don't know, implementation.
don't know.
I say both, right?
So it's like, I'll make the spec, people can still implement it.
And I don't think YAML script will be the only implementation to use, but it will be theone that will evolve and be the most correct with YAML for all languages.
(01:00:17):
So I think that's really important.
Even without the scripting, there's another, there's, can turn off all the scriptingcapabilities of YAML script.
You know, obviously.
You don't want to expose some of the things that it can do, like make web calls, databasecalls, whatever, or access file systems or parts of file systems.
(01:00:41):
Yeah, that's all configurable by either by both the user and by the consumers.
Well, Engy, thank you so much for coming on and telling us all about the past, present,and future of YAML.
I think it's been uh fascinating not even realizing that most of my career has beenevolved around YAML in some way, because I was writing Ansible pretty early on.
(01:01:06):
Even SALTs for config management was YAML based.
And while there was always this Python or JSON sort of bits and...
uh
other pieces it was all a subset of yaml in some way so that's really fascinating.
It's really rad, I didn't know YAML powered so many things.
(01:01:26):
Yeah, I didn't either to be honest.
mean, like, like when I went to KubeCon and just saw the amount of YAML in the world, justkind of messed with me a bit.
It's a bit crazy.
Where can people find you online?
there a place that you want people to find you online?
(01:01:46):
Absolutely, I'm like always, I can use another friend.
Yannelscript.org slash INGYDOTNET.
We can put that down in the links.
This is the page of all the ways to contact me.
So, it doesn't have my address, but if you ask me, come over and we'll do a barbecue orsomething.
(01:02:11):
Perfect, we'll have it in the show notes.
Thank you so much.
Yeah.
Autumn.
It's been awesome.
And you guys do a really good job of making a podcast flow really well.
really enjoyed doing this with you guys.
Thanks for coming on the show.
Yeah, we're just nerds that are interested and so it's fun.
(01:02:32):
It makes it really easy when there's an interesting topic.
When am I gonna, let's see, when am gonna see you again?
Are you going to, you guys got all the coupons?
Are North America's at least?
only wishes.
ah She doesn't tend to go to the cube cons.
hard and then all my friends have fun without me and then Justin sends me pictures.
It's so, it's, it's cruel.
cruel.
That's cruel, Justin.
(01:02:54):
I will be at KubeCon North America in Atlanta.
Yeah, that's in November.
So I will be there.
Sounds good.
Thank you everyone for listening and we will talk to you again soon.
And.
Popular Podcasts
United States of Kennedy
United States of Kennedy is a podcast about our cultural fascination with the Kennedy dynasty. Every week, hosts Lyra Smith and George Civeris go into one aspect of the Kennedy story.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com