> The first version of CONL used # as a comment token, but I quickly ran into issues. URLs contain #, so my next version...
Every other language has figured this one out as well. Wrap strings in quotation marks.
> That led to a data-model where each value is one of scalar|list|map (Compared to JSON’s null|bool|number|string|object|array, this felt good).
I'm not sure what a "scalar" is in CONL (is it always a string?) but a config file format having fewer types than JSON does not feel good to me. Even JSON's hand-wavy "number" type is problematic (whether "1" is an integer or float or some some other type is implementation-defined). TOML got it right to distinguish integers from floats. TOML got this right.
Those who don't learn their history are doomed to find new and innovative ways to repeat history.
If you're older than 40, you remember that there did exist an aeon, long, long ago, when people did not use data object serialization formats as config files. When config files were written not to be easy to parse, but to make it easier for human beings to configure software. When nobody believed there was one single good way to do everything. When software was written not to aid computers, but to aid humans.
> When config files were written not to be easy to parse, but to make it easier for human beings to configure software
Config files have always been a variant of key-value or section-key-value, except that we used to have ad hoc (and probably buggy, inconsistent, incomplete or all three) rules for quoting; array items separated by a mix of spaces, commas or something else; comments (semicolon, percent, sharp) different for each program. Case sensitivity was also a crap shoot, sometimes different between keys and values.
These days TOML (which more or less just works) just works. I have mixed feelings about YAML but certainly I would not swap it with endless variants of sendmail's m4 madness.
Yeah. I've got no idea what your parent comment is talking about.
> When config files were written not to be easy to parse, but to make it easier for human beings to configure software.
*eyes rolling*. All I can remember is the hundreds of hours I've spent trying to figure out how to configure something in Apache httpd, BIND, iptables, and god forbid, Sendmail!!
Config files were written not to be easy to <anything>. There was no rhyme or reason. Every project had their own bespoke config. All from the whims and fancies of the devs of the project.
Good thing that was all in the past and I had no job and no responsibilities. If software today made configuration like they did 40 years ago, I'd just give up!
I'm sorry, nothing beats KDL in terms of readability and friendliness. I've been using it in personal projects for a while, and it is just so pleasant. I wish it saw way more widespread usage.
Similar to HCL which is way safer and clearer IMHO than all indent based craziness. Its lovely to see default values loaded thanks to some extra spaces. Brackets for the win!
You caveman! Everybody knew that watching config though a 30*80 chars ssh display in black/white should be enough for everybody.
Who needs help displays, validation or even sliders?!
"An INI critique of TOML" this is inspired by was discussed in 2023: https://news.ycombinator.com/item?id=37595766.
It received a lot of criticism, particularly for invoking Postel's law.
Really like the philosophy here. Keeping config formats minimal and text-first (rather than trying to be 'clever' with types or logic) feels underrated these days. CONL looks like it hits a nice sweet spot between human-editable and machine-parseable without drifting into 'just use a programming language' territory.
Author here. Seemed like the least bad of the options.
Being able to comment out sections of a config file easily is a prime use-case; and that really implies using newlines as delimiters, and well, you fall into this trap..
I've also been playing around with a configuration format, for similar reasons, although my approach is to make it easy(enough) to read/parse for both humans and machines.
Any feedback is welcomed, but keep in mind is just a toy project which has only one user in mind(me), no plans to conquer the world or solve the config format problems for all :)
We used hocon at a place that I once worked at and I more or less liked it. It did get a lot of abuse though. I think apple released a configuration language that seemed pretty good for the same things that we used hocon. I think it was pkl or something?
OP has a detailed rationale for going with semicolons. Feel free to counter those points, but you can't just dismiss the thing with a "doesn't look right" without any argument.
The first point is addressed in the article; you don't seem to address OP's counterpoint at all.
I don't get the second point: why is that a problem if a semicolon appears in a comment? From what I understand, comments run until the end of the line, so a second semicolon after the first does nothing.
The problem is when a config value includes a semicolon, and the rest of the line gets ignored unintentionally, especially because strings aren't quoted
I think this is something in some assembly formats too? I remember seeing it once and wondering if maybe that's where the idea of ending lines in C with semicolons came from since at least in the examples I saw in school, a large number of lines had trailing comments with a description of what the operation was doing.
IDA uses ; for comments in its disassembler view, but it looks like C-style // single-line comments and /* comment blocks */ are also accepted by certain tools: https://en.wikibooks.org/wiki/X86_Assembly/Comments
Nice! I share a similar set of thoughts and ideals around configuration languages and I'm working on one as well. Mine has a very similar syntax, so you might be interested! You can find it if you dig through my comments.
Every other language has figured this one out: just support trailing commas. JSON5 supports comments and trailing commas.
https://devblogs.microsoft.com/oldnewthing/20240209-00/?p=10... https://json5.org/
> The first version of CONL used # as a comment token, but I quickly ran into issues. URLs contain #, so my next version...
Every other language has figured this one out as well. Wrap strings in quotation marks.
> That led to a data-model where each value is one of scalar|list|map (Compared to JSON’s null|bool|number|string|object|array, this felt good).
I'm not sure what a "scalar" is in CONL (is it always a string?) but a config file format having fewer types than JSON does not feel good to me. Even JSON's hand-wavy "number" type is problematic (whether "1" is an integer or float or some some other type is implementation-defined). TOML got it right to distinguish integers from floats. TOML got this right.
reply