XMPP Council - 2019-10-09

dwd: https://github.com/xsf/xeps/pull/840 I think you forgot this

(Mentioned above)

Is it Council Meeting Day again?

Time flies fast.

I shall be trying to make it, but AFK a bit.

Hi, I am here.

Hi Link Mauve

I am here too

I also just sent feedback on CS-2020 to the list

Breakfast is almost ready here.

Damn, I'm hungry.

I could eat, too.

that’s not fair.

I’m still so sad for the onions I just cut, and I don’t know why.

because you murdered those onions

jonas’, uh, I haven’t received your email yet.

Ah, now I have.

Oh, good lord. It's time for a meeting.

Link Mauve, yeah, the list server is a bit slow

So.

1) Roll Call

o/

Great, that's quorum.

2) Agenda Bashing

as pep mentioned: https://github.com/xsf/xeps/pull/840

Note pep.'s comment above about #840.

I don’t know of anything else

3) Items for a vote

a) ProtoXEP: Message Retraction https://github.com/xsf/xeps/blob/master/inbox/message-retraction.xml

why not: https://xmpp.org/extensions/inbox/message-retraction.html ?

Because I'm half asleep?

But yes, much better URL.

on-list

on-list

shall handle it together with that message moderation protoxep

On-list too.

I'm +1. But willing to be told I'm wrong later. :-)

reaction, retraction, attach-to-fastening

Ge0rG, Retracting reactions is perfectly fine.

dwd: what about correcting retracted reactions?

We could probaby go on with this forever.

let’s move on.

Did I mention I still have some more AOBs pending?

OK, we'll move to:

b) https://github.com/xsf/xeps/pull/840

I’m not sure if that needs a feature

context: https://mail.jabber.org/pipermail/standards/2019-October/036502.html as it's not mentioned in the PR

Is this suggesting that we has an infinity symbol as a legal value?

dwd, I think so

An "max_items=current + 1 plz" option would be handy

dwd, I assume it's just a placeholder, council can of course bikeshed on what to put as a value :P

I'm not sure whether "∞" is supposed to be a verbatim value for that field, or just some elaborate joke.

Oh, what dwd said.

Is the value to be transmitted by the server or by the client?

Well, I'll -1 this then. Having "-1" or empty seems sane for unlimited, but not novel unicode glyphs.

I'm with dwd here.

Also I don't understand the implied semantics of that value.

> Is the value to be transmitted by the server or by the client? both; depending on context

-1 would be bad imo.

In any case, the text needs to be clearer on what the actual value on the wire for "the server limit" is.

pep., It's really not for Council to do stuff like that.

Here, sorry.

Having it mean arbitrarily-high value wouldn’t be nice.

dwd, you mean it's not for council to provide feedback?

Link Mauve, but it does that when you cast it to uint ;-)

(* on some platforms)

jonas’, so with an example?

jonas’: don't get me started about the semantics of 0198 stanza counter overflow.

Ge0rG, :-)

jonas’: it's NOT funny

pep., Sure, we can provide feedback that a symbol I can't remember how to type isn't going to work.

Anyone else voting?

I'd have thought "" would be better than ∞.

But then so would anything I can type.

I meant to say -1

we need a value that is a distinct "explicitly set to unlimited", which "" maybe is not.

-1 for untypable glyphs in particular.

Ge0rG: It's not actually unlimited, is it? It's 'largest server-allowed value'.

it could be just "-"

"", "-", "unlimited", "max", all are better than "∞", I think.

I could go along with "". That probably works on servers now. I doubt they all mandate an integer value here.

Kev: "∞" is better than ""

this discussion reminds me of: > APL apparently used +.× [as infix matrix multiplication operator], which by combining a multi-character token, confusing attribute-access-like . syntax, and a unicode character, ranks somewhere below U+2603 SNOWMAN on our candidate list.

dwd, servers may easily default to 1 for ""

I wouldn’t trust that

also there is ambiguity between "" and unset

especially PEP implementaitons.

Perhaps. Either way, the proposal is an infinity symbol, so the point it moot.

"max" sounds good to me, but anything typable beats untypable, I think.

dwd: Wrong. Prosody has some limited support for the dataforms datatypes XEP.

4) Outstanding Votes

We have two votes expire today

I'll get to those votes after this meeting.

I know I have outstanding.

The value must be something current servers won’t understand imo, otherwise we’re exposing ourselves to implementation details.

I'm sure others do as well.

5) Next Meeting

Kev, "max" sounds good indeed

could council provide me with some clear direction on how to fix the PR instead of just saying no?

Same time next week?

dwd, +1wwfm

+1W WFM

daniel, I think we did. Don’t use "∞", use some text value instead. In addition, make it clear what is the magic value.

daniel: replace ∞ with "max"?

daniel: also it would be nice to have an explanation of the semantics of that value when sent by the client vs. by the server

daniel, You'll almost certainly need a feature if you're doing something existing servers don't understand.

6) AOB a) Georg wishes to thrash out CS-2020, please review https://github.com/xsf/xeps/pull/841

I've modified the XEP with a new intro text, added some more XEPs

more XEPs to be added are asked for

oh, I forgot about that PR; luckily, judging from the summary, my feedback is still valid

Also the "future work" section

> daniel: also it would be nice to have an explanation of the semantics of that value when sent by the client vs. by the server where. in the label?

daniel, Not now, please.

Ge0rG, I’ll be on-list for that.

Ge0rG, I think everything you've got in #841 looks good.

I wasn’t aware of this PR until now.

jonas’ was so kind to propose some more XEPs for CS-2020

...on list

But it looks nice.

please also comment on jonas’ suggested XEPs

jonas’, Ge0rG - do you have an archive link to the email?

oh, yeah: important detail: I want this XEP to get through Council before the Council re-election, so feedback closes on November 5th

dwd: https://mail.jabber.org/pipermail/standards/2019-October/036515.html

dwd, ^ ✏

Thanks. That was of course for the record and not because I'd completely forgotten it.

dwd, the email I sent 4 minutes before this meeting started? ;)

good thing you didn’t forget about it, that would indicate very bad memory ;)

Oh, OK. So I'd just not seen it yet. :-)

Apparently everybody is on-list to this AOB?

Ge0rG, Do you want a formal vote?

dwd: on the PR? No.

Ge0rG, I think #841 looks good, but I didn’t review the yes/no changes if any because they’re hard to read in the diff.

I'm the owner of the XEP, so I can do whatever I want.

Ge0rG, Indeed you can.

jonas’: it was just a `s/#10005/no/g` kind of change

okay

Ge0rG, then only the feedback I already sent to the list stands ✏

jonas’: it was very good, thank you

you’re welcome :)

I feel slightly inclined to abuse my power to add XEP-0379 to Advanced Mobile.

Ge0rG, I think it all looks good. I'm inclined to agree with Evgeny, too - ditching BOSH sounds like a plan.

dwd: see my response on-list

I would be more inclined to keep BOSH with a note that where possible WSS is better.

(But am not vetoing anything that happens based on either outcome)

Ge0rG, something about tokens ;)

(would be an excellent response to evgeny)

jonas’: something about rejected and/or abandoned protoXEPs

Ge0rG, something about SASL-HT something

Ge0rG, Something about client-key.

Ge0rG, something about Instant Stream Resumption (which is Experimental)

dwd, that’s what I meant

jonas’: are you sure it's not Deferred?

Deferred \subset Experimental in my book

Deprecating BOSH might be a plan, but it sounds like it’s still worth it to have it supported by servers.

Could we deprecate BOSH in clients but not in servers?

Link Mauve: this is exactly what jonas’ suggested in the first mail

Ge0rG, not exactly

dwd: we could in CS-2020

Ge0rG, I said clients can pick one

which isn’t deprecating BOSH

Ge0rG, RIght, what I meant.

oh wait, I said something about phasing out

jonas’: but you also suggested to phase out BOSH ;)

but that was an "maybe even" because what do I know about web

As we're coming toward close, can I ask people to keep an eye on Georg's work on CS-2020, and suggest we revisit this every meeting until it ships?

+1

jonas’, is it 114 you want to remove? I'm curious why. Do you want to push for 225? :p

Anyone want to raise anything else before we close the meeting?

Do we have any data about how usable/blocked WebSocket is in the wild, compared to BOSH?

dwd: yes.

I forgot a tiny little thing.

pep., no, and only removing for Core Server. You can host a Component on a standalone piece of software, no need to have support for this in every server.

We need to LC CS-2020 two weeks in advance to the final vote.

That would mean we have to LC it next week

...latest.

Ge0rG, SGTM

I’d like to call everyone on council ( Kev, dwd, Link Mauve, Ge0rG and me) to try to make this work for once.

I.e. vote on list quickly to let it enter LC

Which is why I'd like to vote for LC now, and use the LC phase to sort out "Future Development"

Same, I will propose changes before next week if I have any.

I don’t think we can technically still vote.

can't we vote in AOB?

Ge0rG, Ah, rather renders my suggestion a waste of time.

I’m +1 on LC after #841 has been included. My feedback shall be counted as LC feedback then.

dwd: your suggestion of revisiting it each meeting? It's sound.

Ge0rG, No, I suggest we Last Call it next week, and this week we ensure we've helped get it to the right shape.

It's still in urgent and dire need of content for "Future Development", though

dwd: well, let's go on with that, then.

But please, let's make sure we're in a position to vote on this within the meeting - otherwise we will run out of time.

Yes.

Sound good to everyone?

Yes.

yes

I have a gut feeling that we will run out of time, so I'll pester everybody delaying the vote with messages.

OK. With that, I think we're done.

If we vote LC now, we still have roughly enough time in the worst case.

Because surely somebody will on-list the LC vote until October 23rd.

7) Ite, Meeting Est

Thanks everyone.

Thanks Dave, thanks everybody else.

Thanks. :)

Ge0rG, I hope not.

On-list.

dwd: ^

Did I do that right?

Don't make me sad, Council comrades.

☭

jonas’: do you want me to pile up more changes into #841?

Ge0rG, asking me as editor?

jonas’: yes

Ge0rG, I have no strong opinion one way or the other.

jonas’: I'd like to have a single revision block for everything between CS-2019 and Final CS-2020.

Ge0rG, if you do, remind me at least 24h before the next meeting about merging it

Ge0rG, that’s not going to be possible

we need to make a release prior to LC

and there will be feedback during LC which needs a separate revision block

Ge0rG, I'd rather we merged and published on a release early/often kind of style.

Does that ask for a "Changes since 2019" sub-section?

Ge0rG, Might nudge people to make more comments that way.

dwd: I'm not talking about not releasing, I'm talking about having a history block that's useful in 2020, and not just in November 2019.

Ge0rG, I don't hugely care about changes. People can figure those out by comparing the docs, they're not radically different in layout etc.

dwd, I don’t agree

strongly disagree even

dwd: using our awesome xep-diff infrastructure? ✏

dwd: as a developer, I want to have all the changes at a glance, so that I can evaluate what needs to be dump to bump my CS level..

figuring out xep changes is a pita indeed

i still would like some clear consultation on how to resolve #840; since this is something that council will have to decide i'm not sure that brute force trying different version only to be told -1 every week is a productive method

daniel, I gave you clear advice which I think will be accepted by all members if I read this meeting correctly.

unless council doesn’t feel that this is important in that case i'll just stop

daniel, do you even read what I’m writing?

jonas’, there was another question

daniel> > daniel: also it would be nice to have an explanation of the semantics of that value when sent by the client vs. by the server where. in the label?

to that my answer is: find a suitable location in the prose of the document and insert it there.

maybe around publish-options or the disco#info example

daniel: I don't much care about the where. It should be readable, so "in the label" is probably not the smartest place to put it

probably publish-options

or that

I don’t care too much either way

what jonas’ said

the thing is that none of the other node configs are explained anywhere

yeah

feels weird to randomly start explaining max

which isn’t great

daniel: PRs welcome.

Personally, I will un--1 it as soon as it's "max" instead of <glyph>, and would accept probably a large number of other typeable strings too

Descriptions desirable but not required in this case, for me.

I would +0 or +1 it without a description, but not without a feature flag

Ge0rG, that’s new.

daniel, ASCII character set value, and probably a feature.

max and feature flag have been added fwiw

jonas’: what's new?

Ge0rG, your requirement

Ge0rG did mention it earlier.

I missed it

Anyway, +1 on the current version (having just seen Daniel's changes).

+1, too

also dwd Bookmarks2 should probably explain what to do when node-config-max is not available as a feature

maybe introduce a magic number :-/

Or just require node-config-max?

daniel: the backticks might be less understandable than "" </bikeshed>

we currently decideded on 128 which ~4 implementations use

I thought that and thought it wasn't worth mentioning :)

(backticks)

but +0 now

but 128 is too small for power users!

I’m going to detach myself from this bikeshedding now.

what is the current number of channels listed on search.jabber.network

let that be our magic number

+1 on #840 now. And FWIW, I'm totally fine with backticks - Markdown, innit?

daniel, 7453

dwd: consistency over markdown.

Ge0rG, I agree. I'd go with a full u8 at least :p

pep.: 256 would force implementations to use at least an int16

I'm missing why

because 256 doesn't fit into an uint8

I didn't say 256 :-°

As for Bookmarks2, I'm not sure this isn't orthogonal. You need the maximum to be set to at least as many bookmarks as you want to store.

pep.: I did

(*trying to escape*)

Ge0rG, What about with compression?

dwd: my gut feeling is that compression will be harder to abuse on mobile, while providing significant benefits

I have only disabled compression in my client because the compressor wasn't thread-safe

I think Dave was suggesting that 256 will fit in a uint8 with compression.

Kev: maybe.

RLE the bits, I reckon. What could go wrong?

it's only one bit.

jonas’: I'd like to group 0245 and 0392 into one category. Is "Message display" too broad / too narrow?

dwd, the 'problem' is that to use Bookmarks2 correctly you really want to put max_items in publish_options; and then you need to put some number in there.

Ge0rG, I disagree on that grouping

Ge0rG, because '392 also makes sense in the roster

and when mulitple implementations disagree on that number you end up having to reconfigure on every publish

daniel, Not if the number is high enough, surely?

yes. but then you need to agree on one

Prosody defaults to 256 fwiw

jonas’: right. That was also a request for a better name.

daniel, Not really. I mean, I see it's simpler if you can just say "max", but I don't see it matters much if the number is smaller.

dwd, publishing with publish-options would reject your publish if your max_items request is different from the one already set on the node.

jonas’, Yes.

Ge0rG, "User handle coloring" maybe?

jonas’, But we know if the node exists at the start, so we're not publishing blindly.

“17:48:52 daniel> we currently decideded on 128 which ~4 implementations use”, I’m awfully close to this limit already myself. :/

dwd, but then you need extra round trips

jonas’, You do, yes. My point is that if the server supports max, that's great. If it doesn't, the client has to do more work, but it'll still function.

jonas’, So I'd see this as a SHOULD at best.

This discussion clearly shows that we are piling workarounds on top of workarounds. Can we step back a step or two and fix the design?

and by "we" I mean somebody else than me ;)

Ge0rG, If you want to ship a free forklift with every client.

so to summarize if a server doesn’t support max i need to download the node config; see if the value currently set is "high enough" and if so omit my own value from publish_options?

that should probably be documented in the XEP so people who are not in the chat right now know that

daniel, Sure. And yes, we should encourage/recommend/etc that `max` is supported and used.

Or just don't do BM2 unless max.

Actually, in the absence of max, what really is the harm in reconfiguring on every push?

At the point you push you already know what 'big enough' is, and can just set it to enough plus a few, or something.

Could even pipline the configure, right?

That's what I was thinking, yes.

also very relevant: https://mail.jabber.org/pipermail/standards/2019-October/036506.html

but pubsub is just a cache

> jonas’> Ge0rG, because '392 also makes sense in the roster And in other places.. OMEMO fingerprint has been mentioned in the past, I'm sure there's lots of other places it could be useful

true true ✏