Update EIP-1: Add W3C as a permissible origin

[SamWilsn]

Originally from #6691 by @kdenhartog.

Splitting out so it can be discussed separately.

Changed slightly with my review feedback.

[eth-bot]

File EIPS/eip-1.md

Requires 2 more reviewers from @axic, @gcolvin, @lightclient, @Pandapip1

[xinbenlv]

Great!

[kdenhartog]

I personally prefer mutable references, but understand the purpose in making these immutable via dates and do not object to it. LGTM

[g11tech]

gtg from my side 👍

[Pandapip1]

Would still like the rules regarding the anchor text.

[chaals]

Except it isn't 'Recommendation' status ...

[bumblefudge]

I agree that being more explicit/precise about which stages in the Recommendation track can be linked would be good! "Proposed Recommendation" seems a stable-enough referent for my tastes (given what I've seen of W3C timelines) but maybe the group prefers a hard line?

Also, how do we feel about W3C Registries? I've seen some suspicion of registries per se in other threads but in this case (as upstream specification) I think it would be worth being able to link to them as authoritative...

[chaals]

being more explicit/precise about which stages in the Recommendation track can be linked would be good!

It would indeed.

Having lived through most of the development of W3C process (e.g. I had to produce the first "Candidate Recommencation", which at that time didn't even have a name and wasn't yet a formal part of the Process), I think we could learn some stuff.

Rather than trying to set bright lines about what is an acceptable status it would be better to be clear about what the different things mean, and to require more explanation / justification of a given reference.

That is in fact roughly what W3C's reference policy for Recommendation is now. They tried really hard over many years to maintain something that had stricter rules until it was obvious that it was just a lot of bureaucracy and entanglement holding up work for the sake of not having to think hard about where the links went.

[kdenhartog]

I never realized this didn't make it to Rec status. Right now the main requirement of a link is that it has immutable content and is a stable URL since there's no official way to update links once an EIP has made it to final status. With that in mind, I think it's less about being rec status for W3C and more about having an official date to recognize the version being referenced and being published in TR since once it's hit there it appears to become a stable link to reference.

Side note, I'll double check why that spec never made it to official Rec status since I believe it's been implemented by numerous browser engines at this point.

[SamWilsn]

First and foremost, we want the link to exist forever. Secondly, we want the content of that link to never change.

Ideally, we'd prefer linking to stuff that's ready for production use, but that isn't technically a requirement.

I'm not familiar enough with W3C to know if the regex as written meets those requirements. I think it's good enough? Would appreciate an expert confirming that though.

[chaals]

I believe the answer is "Yes". The regex (without the "REC-" modification I proposed, which is just to match the text that says it has to be a Recommendation) matches documents for which W3C provides a pretty strong guarantee that they won't make changes beyond fixing markup errors, broken links, definite typographical errors, and annotations that note if the document is superseded/obsoleted/... (although they don't guarantee to make such changes.

In particular, it can match drafts that clearly state they should not be cited except as "Work in Progress", and that can and do change in very significant ways. W3C specs more or less always provide a dated URL for a specific version (matching the regex) and a latest version URL that points to whatever is the latest work. given that they do a lot of conformance testing and work hard in the later stages of the Process to ensure that specs match implemented reality, it is probably a good idea to note this somehow in references.