jdev - 2022-08-11

nav, assuming that "that" is what I think it is, then yes

I've left a comment on https://github.com/xsf/xeps/issues/1173 ✏

flow: 👍 – Good point about the versions. Would versioned XMLs + symlink to latest address some of the drawbacks?

nav some yes, but the main problem is that if you refer to a XEP of e.g. version 1.2.1, then you most likely actually want to refer to 1.2 instead

I think ideally XEP versions are coupled with the XEPs namespace (if any), and minor versions of XEP are increment on namespace change

but I am not aware that we have such a policy

ideally it would be easy to establish one

Does the XEP on versioned namespaces say anything on that?

Zash, not sure, but IIRC the other XEP mentions that the XEP version should be raised when the XEP status changes from experimental to stable

which would be not ideal, because we can't couple XEP versions to the XEP state *and* to the XEPs namespace

We can't couple it to the XEP's namespace anyway, because XEPs have several namespaces, deliberately independent.

some XEPs have that

I guess bumping the, probably minor version, of a XEP once any namespace of ther XEP is bumped would work

the point is, as soon as any namespace the XEP uses is bumped, the XEP is not longer the same XEP as before. and if you want stable references, which is always what you want, then you should take that into consideration

and we as bibliography providing entity should ensure that it is possible that this is accounted for

Any change to a namespace forces a version bump of the XEP.

who forces the bump? i may have read something like that in some procedure XEP, but I am not sure. and what kind of version bump are we talking about? major version?

Major is only bumped on status change.

which is kind of odd

But also makes sense.

just because the status changed doesn't mean that it is a different XEP

I don't think I aggree that a major version bump on xep status change is sensible

It follows some kind of SemVer, right?

"You need to implement XEP-0123 version 1.0" "oh, damn, I only implemented XEP-0123 Version 1.0"

"ahh wait, it's the exact same protocol and XEP"

What's even the problem here?

Zash: "some kind"

0 - experimental, 1 - draft, 2 - final

For x

For y, every change is breaking.

For z, no change is breaking.

Zash, I write an paper which references XEP-0123

but without mentioning the XEPs version, a future reader may look at a totally different XEP

take OMEMO as a prime example

"At the time of this writing, the latest version of XEP-0123 is 0.3" in a footnote?

Zash, sure, but then my example above

"You need to implement XEP-0123 version 2.0" "oh, damn, I only implemented XEP-0123 Version 1.0" "ahh wait, it's the exact same protocol and XEP"

(now with fixed versions ;))

+ we don't have stable links to versioned XEPs

which would be kinda nice to have

I think that's just something you have to deal with.

"that" being?

RFC style documents set in stone forever is nice, until a change is needed.

flow, the "living specification" thing

I think it would be possilbe to hit a sweet spot between what we do know and RFC-style-set-in-stone documents

that is, publish every version of the XEP + don't bump versions of XEP status changes

xmpp.org/extensions/0123/version/1.0

and xmpp.org/extensions/0123/namespace/1 for single namespace XEPs

The version changes with typo corrections too.

and xmpp.org/extensions/0123/diff/by-version/1.0/1.3

and of course, get rid of the stupid numbers and use the unique short names of xeps

xmpp.org/extensions/mam/version/1.0 ✏

Zash, that's perfectly fine, versions are cheap

> and of course, get rid of the stupid numbers and use the unique short names of xeps Numbers are convenient.

Same as with RFCs, we tend to remember them by number, for the ones we use more often.

Also, I often short links are best. Both because of quoting in code comments and because they're easier to type.

As for the issue of citations, just assign a unique number (such as a DOI) to each published version and cite that.

s/unique number/unique identifier/

I certainly don't remember XEPs by numbers but by short names. XEP CSI, what number was it again?

I now the numbers of a few XEPs, but I certainly know more XEPs by their short name

I don't think there is a reason to use a number instead of the short name in most places

I think it's a great idea to use shortnames in references. Particularly because XEP numbers are immutable, and that's just a bore in a reference. Mutable things like shortnames work much better.

I don't think of shortnames as mutable, granted, not sure if this is specified