XSF Discussion - 2019-11-03

Why is SXE based on message instead of iq?

Also, why is it entirely ad-hoc instead of reusing e.g. MUC or PubSub primitives?

Why not?

Zash, because the semantics very much match iqs.

SXE?

XEP-0284.

k

Also, why is there both a Jingle initialisation and another initialisation based on messages afterward?

Link Mauve, I guess so that you don't have to ack state changes? Not that I like it, but it could have been the motivation. And also doesn't apply to the initial state offer which really is exactly iq semantics.

Because someone said: "You should use Jingle for that"? 😀

Most likely; I’ll have to go and read the council minutes from back then.

Seems it was only added in 0.0.8, but there is no history from before 2010 in git 🙁

virtual void send( const std::string& message, const std::string& subject, const StanzaExtensionList& sel = StanzaExtensionList() );

And what if I want to send a message without a body and a subject…?

"", ""

\o/

/o\

std::optional?

This is an API I probably can’t modify.

pity

I could add a new one I guess.

Still not sure why I’m doing this in C++ and not in Rust.

because C++ is more awesome than this newfangled mess of "systems programming" languages :)

I’m not doing systems programming though, just adding collaborativeness to a text editor.

Link Mauve: Aaaah, CW those C++ lines plz!!!! NSFMybrain

Sorry, XEP-0382 still isn’t implemented in my client.

Link Mauve, how is your council candidacy going? Not sure which timezone we are using for the deadline, but it's something like today...

Some timezone. :)

I’ll start working on it once I give up on this C++.

You can not become a candidate after the election 😉

I know that!

Link Mauve: just give up on it. I never thought I'd say that, but: RIIR!

RIIC

+Lua

we already have >5 candidates. I am amazed

and two new ones at that

Ge0rG, I wonder if Swiften might not be better than gloox.

But I’ll definitely have to fix the SXE spec too.

Link Mauve: Existing thing already using gloox?

No, it isn’t collaborative yet.

larma, from your application > That said, my criteria for accepting a XEP as experimental would be: […]

I think this misses an important part (which is even mentioned in XEP-0001 IIRC): "It should not duplicate existing protocols."

although that might be implicit in "It should solve a problem."

Yeah, I think it's implicit in there. A duplicate doesn't solve any problems

fair enough

in any case, thanks for running

jonas’: How do you replace / improve on existing things without duplication?

Does XEP-0313 not duplicate some subset of XEP-0136?

Ge0rG, https://i.imgur.com/10k62WA.png RIIR!

larma, ^

RIIR https://www.r-project.org/

:D

pep., I will consider it for version 2.0 😉

hahaha

larma, re your council election again and since you have an interesting view about personal opinions: One thing which has repeatedly come up is the debate between amending existing standards vs. creating new documents to add features (specifically around XEP-0045 and XEP-0060). Do you think this is a matter of opinion or "taste"? If not, what is the objectively correct way? If it is, how would voting on this question work with your goal to avoid personal preferences?

jonas’, tbh, I don't care which, but I want things to move

I think it's similar for larma(?)

jonas’, I just put something related to that topic on my candidacy page. 1. I think both 45 and 60 should probably be Final already. Most of the changes actually applied to them could be done to Final XEPs as well. 2. Many of the changes proposed to 45 and 60 could also be standalone XEPs. 3. Even if things are not properly defined, it can be later clarified that they are not part of that specification and (even opposing) proposals can then be made in new XEPs (for example IQ routing in 45).

To me it's exactly the same if somebody modifies a XEP or recreates a new XEP amending the first one. it's just the form. We as a group could settle on sth if necessary

larma, thanks

pep., yes and no: the rules how things are advanced to Draft should also apply to significant changes and that won't be possible without a new XEP.

larma, I think you make a very good point in that

larma, but in the end, you add a new disco feature anyway

and at that point it’s questionable whether namespace bumps should be needed at all anymore... but that’s a bigger topic, too big for me right now (-> afk)

I would like to get rid of "namespace bumps" and introduce "new namespaces" in new XEPs (which might just look like a namespace bump, i.e. can have only a version number added/increased)

"which might just look like a namespace bump", it's exactly the same to me indeed. and I'm fine with both, as long as things move

Huh?

?

Namespaced mini-extension protocols without breaking everything using the previous version/namespace?

"without breaking everything using the previous version/namespace", we're not magicians

Something something like semantic versioning? You can to some extent add new stuff without replacing the namespace.

You mean "clarify"

No

I mean add new features

Something supporting the previous version won't know about those features, so no problem.

Yeah ok that's fine. I'm mostly talking about breaking stuff

I'm confused

Namespace bumps should in general be barely needed IMO. Most of it can be done using disco and just additional elements

but of course it's so much easier to just break things and do a namespace bump

Sometimes

Well it's easier for the XEP author, not easier for the developers 😉

I don't think that you can make a specific statement, it really depends on the specific case.

err "generic statement"

True.

I am also not against new namespaces, I just feel they belong in a new XEP

A fresh experimental XEP only a few months old where are developers, and that includes implementing developers, are in the same groupchat is probably easier to namespace bump than a year old xep in draft state

If it's easy to namespace bump, it's also easy to just not namespace bump. That's exactly the point

so what's your stance on jingle ft then? should be collect improvements and bump it?

flow, not sure what improvements are you talking about?

IMO Jingle FT should be at least Draft already, so if there are significant changes required they should go in a new XEP

I wouldn't be suprised to find some if I am going to implement it.

Maybe what we should do is bump the namespace of Experimental XEPs even more often, so everyone either gets used to it or gets together and makes it Draft?

larma, so what is the advantage if having a new XEP versus a major overhaul of the existing? (note that i have no strong opinion in that direction, just wanting to hear the rationale behind it)

Zash, would actually be fine as well. Point is: Experimental XEPs should be deployed in wild, so any changes (if they bump namespace or not) shouldn't have any influence on any running systems.

I'm not opposed to having Jingle-FT-2, but it doesn't feel like the question if we should new-xep or bump-the-existing-xep is something that tackles our most important issues

Jingle-ft is fine as for me. Maybe separate XEPs may describe various meta information for it. Like it was done with thumbnails. Jingle-s5b needs fixes

flow, If it's a new protocol it should get a new XEP number so we have a proper name/identifier for the protocol.

Jingle FT Base and a File Metadata XEP?

"My server supports archives." "Which version?" "XEP-0313" isn't really helpful

larma, "XEP-0313 v0.6" tho?

== `urn:xmpp:mam:2`

So we don't define protocols in XEPs anymore but in XEP versions?

and a different version could be a completely different protocol?

That's how it is now

Only for very few XEPs, and I consider this an issue.

So we should define protocols by namespaces instead?

Don't we already, in a way?

On the wire, but not in the documentation.

larma> flow, not sure what improvements are you talking about? That is actually a nice example for one of the major issues we have. We are incredibly bad at collecting and organizing protocol feedback, leave allone incooperating it into a new protocol revision

Speaking of which, do I need to rewrite CS-2020 now in terms of feature namespaces? Which version of MAM should it require? From clients? From servers?

This seems like pain that comes naturally from having protocols under development be implemented and deployed in the wild.

We don't always change namespace for protocol changes of Experimental XEPs

Ge0rG, "From Client? From servers"? What is the issue if the compliance suites that state xep313 (and that implies the current state of xep313)?

Ah yes. Jingle-ft needs some clarification how to properly finish file transfer. For example when we have multiple files in one session. Or when final checksum isn't sent

Probably nobody would care to bump the references namespace if 372 is changes...

larma, I think there is a sensible policy that protocol changes that break interoperability require a namespace bump in place

larma: normally we do bump, except under very closely defined conditions.

> that break interoperability This!

flow: that it's not clear at which moment of time that version is pinned. On January 1st?

Ge0rG, who said something about pinning the version? :)

Ge0rG, I had the feeling we bump if it would break interoperability of live deployments. As per XEP-0001 there shouldn't be live deployments of Experimental XEPs so there shouldn't be namespace bumps. Draft XEPs should and Final XEPs must not break interoperability, so no chance of a namespace bump there.

We _normally_ don't bump if it doesn't affect interoperability, but this is not what larma was asking about?

flow: you implied that. Otherwise, what happens if a namespace is bumped during the CS year?

Ge0rG, larma wrote about "protocol changes". we obviously don't bump for every protocol change

The problem is that we fail to move XEPs to Draft fast enough so that everybody started considering Experimental or even Deferred XEPs as if they were Draft

larma, isn't that the root problem? Experimental deployed in the wild?

LC all the XEPs!

Ge0rG, then the compliance suite refers to the current state of the xep

flow: so you say the CS defines a moving target?

Should the compliance suite even be allowed to reference Experimental XEPs outside the "future stuff" section?

Zash, or, maybe even better, make XEPs IETF style immutable

Zash: should we have a process in place that moves Experimental XEPs forward when they are needed?

flow, I kinda like that, but not enough to rewrite XEP-0001 for it

Ge0rG, XEPs are, to some degree, always a moving target

Ge0rG, IMO CS shouldn't be able to point to Experimental XEPs so that it cannot point to a moving target

*highly moving target

And I don't see the point in pinning XEPs to particular namespace in the compliance suites

larma: that would stall XEP progress even more

What Ge0rG said

Why?

I have the feeling because XEPs have a hard time transitioning from 'experimental' to 'draft'

larma: developers have hardly the time to implement one version of the most relevant XEPs. How are they supposed to implement multiple versions?

Why should they be required to implement multiple versions just because we move experimental to draft?

Can we please focus on realistic ideas, given the overall xsf time budget?

Plz no

Ge0rG, I too, fail to see where the "multiple versions" now comes from

flow, ns bump transitions?

Zash, yes, but why do developers have to implement multiple namespaces if the compliance suites are not allowed to point to experimental XEPs?

If we have one version in the CS and then the XEP is bumped, nobody will implement the new one.

Speaking of time budget, I'm out.

Ge0rG, You mean if there is one version in CS and there is a new Experimental XEP replacing it, it's not going to be widely implemented? Yes that's the whole idea of Experimental XEPs. As soon as it becomes Draft, the (next) CS can point to the new version. No need to implement two versions at the same time.

larma: and then next year everybody already has implemented the changes, disables the old version and enables the new one at midnight, January 1st?😉

Flag days?

Ge0rG, well transitions are certainly a problem when breaking backwards compatibility, but that isn't any difference to the current situation. I do agree we should try to not namespace bump as long as possible, a new XEP doesn't imply a new namespace as it can also build on top of an existing XEP.

larma: just because you call it "new XEP" it doesn't automatically give people the time to implement it and maintain it aside to the old one

Reality is, some developers will keep multiple versions in parallel, sometimes in different modules with different feature sets. Other will just bump in the least uncomfortable version release

That would break CS compatibility with pinned namespaces

I don't see the point you want to make here

“Zash> isn't that the root problem? Experimental deployed in the wild?” not put like this but yes. The problem is that we live stuff lingering as experimental. So yes, LC more things

People wouldn't adopt new versions, meaning that the CS won't be updated in a meaningful way

“Zash> I kinda like that, but not enough to rewrite XEP-0001 for it“, what's wrong about rewriting 0001?

pep., thanks for volonteering

People are afraid of change and that's the one thing I want to stop, it's quite annoying

Zash, I'm sure I'm not alone

And can we stop this "blame game", or pushing stuff onto each other game

I'm sure we can manage to do stuff collectively and not just as one person at a time

Yes we're all volonteers, so what

Welcome to volonteer based organizations.

Getting volonteers to do things is Hard.

I believe we can do better than pushing stuff onto each other because

Ge0rG, The CS isn't supposed to be "This is what everyone else implements" but "This is what everyone should implement". Thus you can update the CS to the new version even without everyone implementing if you think it should be. You can also say something "you are still CS compliant this year if you implement the old standard, but next year we will probably require the new one". However I wouldn't call a client or server Advanced CS2020 compliant that only implements mam:tmp, it should probably be at least mam:1

"This is what everyone should implement" -> "This is what Council suggests everyone should implement"

Zash, "Getting volonteers to do things is Hard." especially if they ask to help and receive the answer "it's none of your business"

(not meaning you)

Is that something someone said here?

It did happen in the past yes.

pep., FWIW, nothing wrong with rewriting XEP 0001, just not a current priority for me.

Zash, what I meant by that is, if we have to rewrite 0001, then so be it. I won't be afraid of it

I'm annoyed by the "thou shalt not change XEP-xxxx" and all the fearmongering around changing things

We cannot go ahead without changing things. XMPP would be dead already if it were not possible to do so. (Some still say it still is, but that's an opinion I'm happy to discard. Lots of changes happened in the previous years and change will come again)

pep.: people even dared to change 0045

pep., just do it!

Somewhere in the previous 300 messages someone mentioned me. If it's important, drop me a mail please.

XMPP honestly feels like it's been reviving - especially with omemo being supported by more things now.

I wouldn't especially name OMEMO, but ok :)

That definitely played in the acceptance by users

it's what got my circle using it instead of things like Signal, at least - encrypted group chats combined with being able to pick a client according to individual preferences is a big deal.

I'd like to avoid security discussions right now when we're talking about process

sorry!

(Not that I want to prevent you from talking about this, just that this topic is a minefield, and we're already discussing another)

well, we were..

pep., so, where's your board|council application? 🙂

I'm procrastinating it. Give me a sec..

Wait really? XEP-0081: Jabber MIME Type https://xmpp.org/extensions/xep-0081.xml

And https://tools.ietf.org/html/rfc3923#section-10

I was looking for this very thing the other day

haha what

https://xmpp.org/extensions/xep-0104.xml

This is shameful

PREFIXED ATTRIBUTES!

there is the precedent I needed!!

hmm the prefix in xep104 seems unnecessary

just like the UTF-8 requirement in rfc3923 § 10

but: yeah, using prefixes /o/ \o\ /o/

Oh no

although I'd rather use them only when there is no alternative

they can be used to reduce traffic, but (in most cases) not in an RFC-compliant way.

RFC-compliant way?

RFC says something about avoiding prefixes?

https://xmpp.org/rfcs/rfc6120.html#streams-ns-declarations sounds like a relevant heading

> It is conventional in the XMPP community for implementations to not generate namespace prefixes for elements that are qualified by extended namespaces https://xmpp.org/rfcs/rfc6120.html#stanzas-extended

jonas’: so, we-dont-do-that-here.jpg

> Routing entities (typically servers) SHOULD try to maintain prefixes when serializing XML stanzas for processing, but receiving entities MUST NOT depend on the prefix strings to have any particular value E.g. Prosody will likely spit out such things as ns1:foo, ns2:bar etc

Zash, does any of this talk about attributes?

Not directly from what I can see

Why wouldn't that be the same as for elements?

here we go

Link Mauve, your application.