Some dfns are case-sensitive

[annevk]

We started introducing dfns in HTML for which casing matters. E.g., StructuredClone. @domenic noticed a problem with bikeshed lowercasing dfns here: whatwg/url@7b40216#commitcomment-16540044. Can that be changed somehow?

[domenic]

What's weird is that it doesn't lowercase everything. Maybe just dfn-type definitions?

[tabatkins]

It's dfn-types and a few more: https://github.com/tabatkins/bikeshed/blob/master/bikeshed/config.py#L94

Can you elaborate more on what you're defining here? "dfn" types definitions are generally English, where casing can vary unpredictably. (In particular, a term appearing at the start of a sentence will be uppercased, and shouldn't require extra effort to link.) From the diff, it looks like HTML switched from an English-style name to a pseudocodish CappedAlgorithmName.

Maybe this means I should introduce an "algorithm" definition type, for this kind of "coding in English" function thing?

(Bonus: if I can tie this into the existing "algorithm" concept Bikeshed has, for scoping the use-checking of <var>, you'll get better behavior for free!)

(I don't want to change the behavior of dfn-type definitions here, even to something innocuous like "just lowercase during matching, preserve the original case when displaying", because you can start a sentence with a <dfn>, where the fact that it begins with an uppercase isn't an important part of the term, and seeing a capital in the index would be distracting.)

[domenic]

A new dfn type sounds pretty reasonable to me. We'd add some data attribute to the dfn for Bikeshed to become aware of it? Alternately I guess a data attribute that says "preserve case", in case there are other non-algorithm use cases where this is important.

And yeah, it makes sense to lowercase things to allow people to use <dfn>s at the beginning of sentences.

[annevk]

It gets a little weird if your dfn has abbreviations and such in it. You can also use lt if you put <dfn> as the first thing.

But I'm okay with trying to make algorithms work. But let's call it "abstract operation" then, shortened "ao" by the ECMAScript folks.

[tabatkins]

Ok, added an "abstract-op" type.

I'm planning on turning the "dfn" type into a supertype, encompassing a new "term" type (the old "dfn", for ordinary English terms) and "abstract-op". That way linking with a plain <a> (or the [=foo=] shorthand) will catch abstract ops too.

[tabatkins]

Never mind, changing the link type fucks things up pretty deeply, and it turns out that mixing types with different capitalization rules into a supertype is not a good idea.