Human-Oriented Markup Language
huml.io
huml.io
25 minute talk at https://www.youtube.com/live/AUrPdOZNsX8?feature=shared&t=13... (starts 3h:52)
The talk proposal is at https://fossunited.org/c/indiafoss/2025/cfp/arsnhack6n
And it primarily tries to avoid YAML horrors : https://noyaml.com/ and https://ruudvanasseldonk.com/2023/01/11/the-yaml-document-fr....
Can someone explain to me what is so horrible about curly braces that we need a whole host of "human-friendly" configuration languages with nontrivial parsing just to get around them?
s/JSON/YAML/g
s/curly braces/whitespace/gWith JSON it's fairly easy for me to know if I want to end my structure as well as the structure containing it, I can just type }}, and add the next element.
With whitespace you have to keep track of HOW MUCH whitespace, and trust me once you've got people who are entirely inconsistent with how much whitespace they use it becomes a huge PITA.
I find that paste/format on copy/paste with YAML can be kind of troublesome too though... I have an exception in my vscode settings file for that format, my default is format on paste and save.
< date="today" id="978"/>
I could have used this many times.
It seems trivial, but replacing scope delimiters with a per-line signifier (i.e. indent) makes the scope of each line self-contained and sidesteps that discussion.
Is that worth YAYAML (Yet another YAML)? I don't know. But I certainly get the desire to skip the discussion :)
The following would have parseable structure, but would be free of visual noise.
Title: Markup languages: decades of going in the wrong direction
Keywords: hypertext,
delimiters,
ˋ, ", \Control characters are invisible, using them means changing text editors to display them. They are also, outside the usual ones, hard to type. ASCII ones have Ctrl combos, but editors used those for other things.
Also, what is the difference between using some new character to start block and "{" or "\n"? Why have new thing to indicate new level when have space and tab?
Yes. Change of paradigm does require change of tooling. If some legacy tool doesn’t support new format, it’s not a good reason not to use new technology - either tool evolves or a replacement emerges and typing won’t be a problem. Classic formatting commands from rich text (Ctrl-B etc) can be repurposed, for example.
>Also, what is the difference between using some new character to start block and "{" or "\n"?
Any such delimiter has other use in text. Dual use means extra ceremony with escaping and extra complexity. Whitespace as a delimiter has especially bad UX, because most editors don’t understand the semantics and it is very to make mistakes.
It depends, if you allow nesting you also need escaping.
They are not used in editors that I have ever seen.
They are theoretically accessible as ctrl-/, ctrl-], ctrl-^, ctrl_
Editors absolutely should be able to show them using syntax highlighting.
Probably most of the first 32 can be repurposed.
Allowing only valid JavaScript identifiers to be unescaped keys is also a bit quirky (You have to quote reserved keywords).
But they will never change any of that because it would break JavaScript compatibility.
I shouldn't have to care about what the type of the value is when writing out effectively YAML. This double-colon feature will do nothing but lead to bug reports from people confused as to why their document is invalid.
For example `x: 3` would be equivalent to `"x": 3` in JSON, but `x:: 3` is equivalent to `"x": [3]`
(One could question how human friendly it is to call lists and dicts "vectors" though...)
props:: mime_type: "text/html", encoding: "gzip" # Inline dict. props: mime_type: "text/html", encoding: "gzip" # Inline dict. props: [mime: [html, encoding:[gzip]]]
There are many different humans. I definitely like the idea to separate “:: vs :”.
Also notice what the commenter above haven't done (yet, maybe they will?): done a full "forest level" comparison of all the trade-offs between the current HUML specification and ... what is your alternative proposal, exactly?
Based on my experience, I would guess that most people who design a language (and written a parser for it) for the first time will: (1) be surprised at how quickly design decisions snowball and lead to unexpected places; (ii) discover just how entangled design choices really are; (iii) will give up on trying to please everyone.
In my view, a language designer does really well to describe one's motivations, goals, tradeoffs, decisions, and then live with what you make, because... (a) making something real and useful is rad and (b) any language you make will probably have some weird stank you can't seem to get rid of.
https://stackoverflow.com/questions/6968366/if-yaml-aint-mar...
On the other hand, and to your points, Relax/NG (both its XML and simplified syntaxes) is a beautiful successful example of wisely and collaboratively designing a new clean powerful system with the deep understanding of what a markup language really is (James Clark was deeply involved with many SGML and XML standards and implementations), and full cognizance of the strengths and weaknesses of other systems you're trying to replace (SGML DTDs, XML Schemas, TREX, RELAX, XDuce, and other experimental XML schema languages).
https://en.wikipedia.org/wiki/RELAX_NG
>In computing, RELAX NG (REgular LAnguage for XML Next Generation) is a schema language for XML—a RELAX NG schema specifies a pattern for the structure and content of an XML document. A RELAX NG schema is itself an XML document but RELAX NG also offers a popular compact, non-XML syntax.[1] Compared to other XML schema languages RELAX NG is considered relatively simple.
RELAX NG Compact Syntax
https://www.oasis-open.org/committees/relax-ng/compact-20021...
https://relaxng.org/jclark/design.html
The Design of RELAX NG
James Clark (jjc@thaiopensource.com)
Abstract: RELAX NG is a new schema language for XML. This paper discusses various aspects of the design of RELAX NG including the treatment of attributes, datatyping, mixed content, unordered content namespaces, cross-references and modularity. [...]
>Composability
>RELAX NG is designed to be highly composable. A schema language (or indeed a programming language) provides a number of atomic objects and a number of methods of composition. The methods of composition can be used to combine atomic objects into compound objects which can in turn be composed into further compound objects. The composability of the language is the degree to which the various methods of composition can be applied uniformly to all the various objects of the language, both atomic and compound. For example, RELAX NG provides a choice element that can be applied uniformly to elements, attributes, datatypes and enumerated values. This is not mere syntactic overloading. The choice element has a single uniform semantic in all these cases and can have a single implementation. Another example is the grammar element, which is the container for definitions. The grammar element is just another pattern and can be composed in just the same way as other patterns. Composability improves ease of learning and ease of use. Composability also tends to improve the ratio between complexity and power: for a given amount of complexity, a more composable language will be more powerful than a less composable one. [...]
>XML syntax
RELAX NG uses XML instance syntax to express schemas. Although this makes for a rather verbose schema language, it has some major advantages. Since a user of an XML schema language must necessarily already learn XML instance syntax, using XML instance syntax for the schema language reduces the learning burden on a schema user. It also allows XML tools and technologies to be applied to the schema. For example, a schema can be used to specify the syntax of the schema language. Another important benefit of XML syntax is extensibility. RELAX NG has an open syntax that allows the RELAX NG defined elements and attributes to be annotated with elements and attributes from other namespaces. RELAX NG DTD Compatibility [12] uses this annotation mechanism to extend RELAX NG with a mechanism for declaring default values for attributes. RelaxNGCC [23] uses this annotation mechanism to allow users to embed Java code in RELAX NG schemas, which gets executed as an XML document is parsed against the schema. An unofficial non-XML syntax for RELAX NG has also been developed [8]. The non-XML syntax can be used for authoring RELAX NG schemas by hand and can then be transformed into the standard RELAX NG XML syntax for interchange. [...]
I'm directly illustrating your point that "a language designer does really well to describe one's motivations, goals, tradeoffs, decisions, and then live with what you make", with one negative example of YAML spectacularly failing, and another positive example of Relax/NG spectacularly succeeding.
Because if all I do is beat a dead horse by ragging on YAML without suggesting any counter examples, I'm just pointlessly whining.
The YAML developers were so ignorant of and incurious about past and existing technologies that they were under the mistaken impression that what they designed was a "markup language", so pathologically that two out of the four letters of its name are incorrect, and they had to change what the other two letters stood for retroactively after they finally figured it out. And the human interface -- the syntax of YAML itself -- suffered deeply because of that ignorance (as do its millions of users and apps and LLMs that still have to deal with those bad uninformed design decisions).
The Relax/NG developers knew what the hell they were doing because they had a long track record of working with real world markup language standards, and they eloquently and precisely described their motivations. They even KNEW what a markup language was, unlike the original YAML designers.
They deeply understood not only the history but the cutting edge competing technologies of the time, and Relax/NG greatly benefited because of it. They expertly performed, as you suggest, ''a full "forest level" comparison of all the trade-offs between the current'' RELAX/NG specification and many other schema languages past and present.
Bristling when somebody enthusiastically tells you about old technologies you should look at before designing new technologies is just self imposed ignorance, not a good look. Cultivating ignorance instead of learning from the past is what got the YAML designers in trouble, too.
So no, I'm not a Relax/NG evangelist, I have no stake in it, and I haven't used it for well over 15 years. And I've also used XML Schemas, which is another negative example of terrible language design, and why I appreciate Relax/NG so much. When I was learning and using Relax/NG, I read a lot of the papers and code on James Clark's web site, as well as the archives of the Relax/NG design group discussion (which were fascinating, if you're interested in user friendly language design and markup languages and schemas, which I am, and which this discussion is about).
Of course I've written before, responding to other people's comments, about how Relax/NG compares to XML Schemas, the history of markup and schema and language design in general, and how amusing it is to compare the length of James Clark's Haskel implementation of Relax/NG with the length of his Java implementation (although all arguments about how terrible Java is these days can be instantly won by mentioning the word "lawnmower" before getting into the weeds of thorny technical issues):
https://news.ycombinator.com/item?id=28668752
But I'm not trying to convince you to use Relax/NG, I'm just suggesting you might learn from its design, and check out James Clark's design document as a shining example of how enlightened language designers should express their goals and intentions just like you said.
I'm sorry you can't or won't see it, but what I learned and shared about Relax/NG applies directly to this discussion about YAML and HOML's reaction to it, insofar as Relax/NG has well designed user friendly human readable and writable compact syntax, which is perfectly compatible with its clumsy verbose XML syntax.
James Clark's paper, The Design of Relax/NG, is a perfectly on point illustrative example, directly addressing your comment about how language designers should "describe one's motivations, goals, tradeoffs, decisions, and then live with what you make".
https://relaxng.org/jclark/design.html
Did you bother reading or even skimming that excellent paper which exemplifies what you called for, before accusing me of not responding to your comment and hijacking your thread? Geez, sorry, dude. Read the paper, then you'll understand.
You can lead a horse to water...
Rather than ignoring your comment, I was open about how it came across. Some people don’t like that level of directness. It is hard to take feedback gracefully, I get it. A lot of people prefer to counterpunch.
Also, I think it’s probably a stretch to hope a person will read or skim through the quantity of words you’ve quoted at length and linked. Think about the readers. Whether you like it or not, you have convince people that it’s worth their time, one way or another.
To use your metaphor, you didn’t lead the horse to the water. The horse didn’t see water; it saw a pile of undrinkable words.
What an awful phrase. Leading a horse to water is about giving an opportunity to drink. If you lead a horse to water and it doesn’t drink, it likely means it’s not thirsty! You don’t yell at the horse for not drinking it.
All this said, I do sincerely appreciate you taking the time to reply.
I think it’s a neat improvement.
> Provide as few ways as possible—preferably one—of representing something.
Very Pythonic. Especially since representing a dict already has 2 ways, on the first page!
And lists.
> Pythonic
Pythonic in the way that Python's 'There should be one...' is expressed through the existence of tuples, named tuples, dataclasses, regular classes and attrs (not part of the standard library but it seems to be as much of a goto as requests is)? ;)
I love Python to death, but will readily admit that in no way is there “one obvious way” to do things.
But why double the character instead of picking another one... plenty of other non-alphanumeric characters to chose from.
If
numbers: 1, 2, 3
numbers: 1
Tokens are inseparable from human instincts of single = less, double = more and the corresponding emotions: single=less=easier=quicker, double=more=complicated=longer=difficult
If you are not emphasizing the single token as them most common, it's going to cause confusion.
Not sure which one we have here.
Or restrictions like "only one space allowed after : before a value"
TRUTH!
Or, expressed in YAML 1.1 [1]
y
Y
yes
Yes
YES
true
True
TRUE
on
On
ON
trueWeeks of my life have been reclaimed thanks to TypeScripts LSP.
All configuration languages suck in Monaco, so much so that I would rather use C# or TS to generate a config with docs in the tooltips, red squiggles under my typos, and location-aware autocomplete.
We don't even need HCL (offers little value), YAML (has too many features making it inherently less secure), XML (is a mess with too many features and too much verbosity), or INI (insufficient features) except for existing and historical reasons.
In cases where a single file is attempted to be managed by multiple, competing interests, it's probably better to split the file into multiple, modular/include files like *.d/* so there would be less chances for merge conflicts.
And then there's /{proc,sys}-like configuration with one "value" per file where the path forms the heirarchy.
Please. Just. Stop reinventing wheels that were already round enough.
Term not found.
JSON prevents this by explicitly closing lists and dicts. Though you do end up with Ladders to Heaven of `}`'s and `]`'s where you have no goddamn clue what is actually being closed out.
XML prevents this by explictly stating what element was just closed. Deeply nested XML, provided the elements don't have some 300 character set of attributes all jammed into the opening element, is vastly easier to read than YAML.
I might understand this objection for YAML, but for python, it's very clear: a block ends whenever the content on the next line is indented by four fewer spaces. (And lists and dicts are explicitly closed with brackets)
If you're talking about matching block starts to ends, then that's just as bad in json, if not worse; without an editor that does brace coloring, it's almost impossible to eyeball which brace goes to which block without indenting (which is the whole point of significant whitespace).
> Deeply nested XML, provided the elements don't have some 300 character set of attributes all jammed into the opening element, is vastly easier to read than YAML.
Not really. More often than not, tags are visual noise that make it harder to scan for the actual content.
I have to read a lot of deeply nested XML for part of my job, and I wrote a parser to display it as YAML precisely because it's more readable. If your data is so convoluted that you need xml tags to make sense of it, that's usually a sign you need to either refactor it, or provide a proper display tool.
To match the start of a Python block (with possibly other blocks nested within it) to its end: start at the (non-whitespace) column where the opening line begins, and go straight down until you find another line that starts in that column (or EOF).
Or you use an editor (or plugin) that just knows how to do this, just as you would with the balanced brackets. cf. https://vi.stackexchange.com/questions/7262 .
You can also just... not nest so deeply, or write so long. My functions rarely have more than ten lines are are usually not even inside classes; and I don't generally need more than three levels of indentation (that gets you, for example, a test within a loop within a function — which might well be better done with a comprehension anyway).
> XML prevents this by explictly stating what element was just closed.
Only for elements that cannot, even indirectly, contain other elements of the same element name. And for this limited safeguard, you double up every element name even in cases where it would otherwise be completely clear (which contributes to visual noise as the sibling comment points out). Of course, XML is designed for things that inherently need to be deeply nested. Your code usually doesn't.
https://en.wikipedia.org/wiki/Turing_(programming_language) tried the same idea. It didn't catch on.
Way back in like 2012-2013 I was putting together workflows in ESRI ArcGIS.
The options for scripting were either VBScript or Python. Python was the better choice (AFAIR the UI would show what you were doing in it as python, which made it easy to take a manual flow you developed and automate.)
However, the editor window for scripts in the UI, used the default windows dialog font configured for the user. [0] I wound up having to put together a workflow where I'd have to copy-paste back and forth from a text editor, which was fairly unproductive.
[0] - No, changing the dialog font to fixed width was not a great solution, because then other apps suddenly had terrible UX due to their design....
The ones who read the XML file understood immediately. Why ? Because XML is self-documenting and the repetition via clear de-limitation extensively aids human memory.
The one who read the JSON file just glazed over it and needed to refer back to the document and re-read several times.
I urge all the XML naysayers and XML haters to please carry out this practical test with their colleagues. Please choose a good level of complexity and nesting and please make use of XML attributes for scalar values.
XML is SO FAR ahead in readability and grokking that it is not funny.
I really don't get it; the whole point of computing is to make stuff rigorous and precise so that you don't have to run a whole guesstimate translation layer as you need to with humans.