today in "I'm glad XML and its ilk are a dying breed": what is the fucking point of having xmlns:whatever="url" for like three type of metadata you see every time someone uses e.g. RDF if there isn't even a validation tool that can check well-formedness by recursing into those URLs?!
JSON Schema is not especially good as a format (this is mostly but not entirely down to JSON's questionable qualities) but every document gets to have a $schema that refers to its machine-readable schema (which, themselves, can reference other schemas via $ref or $dynamicRef). problem fucking solved
I would be a lot less annoyed about this if writing RDF metadata didn't feel like a particularly esoteric form of sadomasochism
Oct 22, 2026, 7:32 PM@whitequark I love XML and it pains me that you are correct
@mcc @whitequark Yeah there was the whole thing that they were URNs rather than URLs and so they couldn't be reliably used as a resource locator, so it was only some companies actually made them resolvable to an Xschema, which could be useful. You would have all the relevant schemas pre-loaded tho, usually, so you wouldn't be doing downloads. (I was doing W3C standards back then - but memory may be wonky)
Oct 22, 2026, 7:34 PM@whitequark wait there isnt a tool thst does that?!?!
i never really got into XML, but i always assumed that well, all that "we specified that thing here" was actually being used and wasnt just larping "we are well specified"
@4censord YOU WOULD FUCKING HAVE THOUGHT.
Oct 22, 2026, 7:36 PM@whitequark the trouble is that that the value you use with xmlns there isn't a url that points to something, it's just a unique identifier
Oct 22, 2026, 7:39 PM@whitequark when they say namespace they mean just that
Oct 22, 2026, 7:40 PM@whitequark you need some kind of schema (e.g. dtd or xsl) attached to the namespace to actually have information about possible elements to validate against
Oct 22, 2026, 7:42 PM@whitequark s/xsl/xsd/
@kitten_tech @rakslice except we had those exploits (XXE) anyway! they were just in DTD expansion rather than lookups through xmlns
this is the worst of both options: you don't get useful functionality and you get the exploits
@rakslice thats fucking stupid. Make a good specification instead
Oct 22, 2026, 8:17 PM@whitequark don't forget XML is from before constant bigbrother24/7 internet connection. In this context it makes a lot of sense. I'm kinda worried about today's assumptions of constant internet access tbh. And with the internet being pretty much in the hands of a few companies... You see where it goes...
@rakslice
Oct 22, 2026, 9:07 PM@whitequark technically yes, but it still assumes connection by design. But that doesn't change the relevant part that XML doesn't assume network because it wasn't something at the time. And breaking compatibility is always a huge pita. Just look at how many websites are still locked on vulnerable php versions because the latter ones are not compatible. I'd rather have clunky specs than having to change code every few years to stay compatible. And that's what would have happened with XML parsers if it did that.
@rakslice
@rakslice @whitequark At least with SNMP it refers to an actual schema, even if actually finding and downloading said schema was left as an exercise for the student.
@whitequark @rakslice Ah, maybe it is.
I think I misinterpreted the issue as lack of a validation tool, rather than discoverability of where the schema/MIB might be found.
Then yes, same problem. Sorry.
@whitequark Oh, writing RDF-XML is *horrible*. I strongly recommend Turtle (or TriG, if you need named graphs).
Or N-Triples, if you want to do bulk operations with line-based text tools. You can do a lot with N-Triples and grep and sed...
@darkling that looks at least as upsetting to the senses as the raw RDF-XML
@whitequark It's much more compact and a lot more readable, I find. When I'm hand-wrangling RDF, it's my preferred format.
Turtle, N-Triples, RDF-XML, JSON-LD, in that order. JSON-LD is the work of the devil.
@whitequark @darkling I attended a DARPA Semantic Web conference back in 2003 at which I saw Tim Berners-Lee give a presentation about some CLI tool for working with one of these triple formats. I thought all of that stuff fell out of use after the Semantic Web hype died out.
@tedmielczarek @whitequark I was using semweb stuff in my last job. It was specifically why they hired me, as I started using RDF and friends slightly before that talk you attended. :)
@whitequark Well, at least I'm glad XML and its ilk are a dying breed.