XSF Discussion - 2020-03-02

larma, at some point I wanted to go through all deferred XEPs and move them either to obsolete or to experimental, but Council couldn’t agree to a process.

Link Mauve: The process is there, and AFAIK all it needs is someone to drive it (does not need to be in council)

all you need is to agree to shepherd all the XEPs you propose

nobody stops you from doing that

(actually we're already LC and CFE-ing a bunch of XEPs at the moment, so *right now* I'm going to say "wait until we're done with the shortlist")

larma> How do people feel about enabling the Deferred check box by default on https://xmpp.org/extensions/ +1

+1

+1

jonas’, yeah indeed.

Why is there no registry for ad hoc command nodes?

And they are also not namespaced, so doesn't that imply that node names are opaque strings?

XEPs relying on ad hoc command nodes having a specified name are therefor kinda broken by design

larma, there is a XEP which defines a bunch of ad-hoc commands with specific names

133 does, but it is only best practice

so they are not well defined, just a suggestion

it’s informational track, which is good enough

it doesn’t define wire protocol, just usage of an existing protocol

yeah, I guess that's fine for 133

and they’re namespaced, just in a weird way (with URL#suffix instead of the more well-defined {URI}suffix)

but yeah, in the Ad-Hoc XEP itself they’re opaque

also kind of the point of ad-hoc commands to be, well, ad-hoc

but there are others in experimental that define commands with a given name

a registry would probably be good, but you know the state of the registries

like 401

401 is using URL#suffix as well

yeah well my point is that any XEP using them would kind of risk name clashes with other names. As according to 50 it is not supposed to be a well defined name and be completely opaque (basically only used as an ID in response to the command list)

according to xep 50 it would be perfectly fine to use UUIDs instead of names

or unicode emoji...

yeah, therefor the fact that you are using URL#suffix is worth nothing, because it is opaque and thus I could use the same name just as a selection of random chars

a perfectly valid server implementation could generate random node names on every server restart

it could, but then it would also be useless

why?

according to the spec, you are requesting the list of commands first and the node name is not even supposed to be displayed to users

because automated entities will not be able to interact with the commands

yaxim will rely on 0401 having that exact node name for invitation generation to work

jonas’, automated entities can still automate with other entities. If they don't want to rely on the command listing before they would kind of have to agree on a set of commands out of band before

like in XEP-0133?

it's just a suggestion

and frankly I believe nobody relies on it client side

you can just do command listing before displaying administration commands

And the randomly-picking-new-name-server would still work perfectly fine with everyone doing that

a lot of things would work perfectly fine for some use-cases

that’s a pointless statement to make

you can ignore a lot of SHOULDs before you run into troubles

My point is, I read XEP-0050 as it was considered to happen that way

xep-0050 maybe, but xep-0401 most definitely not

pubsub node names are also opaque

but people would start to complain if you put OMEMO keys in a randomly named node

then again my point is that xep-0401 is abusing another xep out of it's intended usecases

(or avatars for that matter)

same situation really

So issue with use of 0050 is fine because we already have issues with use of 0060?

no, because it’s not an issue

if one specification tells clients to not assign any meaning to a value and another specification inherits on that and says "oh, in this case, you can assume that the thing you want has name X", then that’s perfectly valid

a new abuse case by 0401?

that’s the whole point of not assigning any meaning to a string in the base spec

if it doesn't have any meaning in the base spec, how can another spec imply any meaning in it?

easy. you write it in the text.

https://xmpp.org/registrar/nodes.html < pubsub nodes are supposed to be registered at least

that’s working well: Last Updated: 2004-10-11

and yes, ad-hoc should also have a registry, I’m agreeing on that one

well the only node that is errornously not registered is 0084 user avatar

and all the nodes from '133

ah, you’re with PubSub, not with Ad-Hoc, sorry

but don’t they share the same namespace?

what?

ad-hoc and pubsub and disco

what do you mean by sharing namespace

They all use nodes

^

ad hoc node names are not the same as service disco node names

sharing the namespace: undefined things happen if you have two things with the same name

From XEP-0030

larma, good then

we just need a separate registry

and a document like the FORM_TYPE thing which defines that it shall henceforth be used ✏

that would work, so who is going to file the PR against 0050?

(btw, this stems from flows suggestion to use ad hoc commands on the mailing list)

haven’t read up on today’s stuff yet

larma, the more interesting question is: who is going to fix the registries?

there’s zero point in adding a registry if we lack the tooling to maintain it

what do you mean with tooling? it's not like we advance tons of XEPs to Draft such that we really need tooling around it

larma, there is no way to get changes to the registries on the website at the moment

and by no way, I mean that you’d have to have iteam manually upload html files, which is Not Gonna Happen

ok, that's bad indeed

yeah.

best case, the XMPP Registrar Considerations sections in XEPs would be machine readable and we would automatically generate the registry list from XEPs (plus additional predefined things where needed).

that’s already mostly machine readable (with soem heuristics)

but it wouldn’t catch cases like '84 which simply miss a submission to a registry

so we’d also need tooling to detect such cases

but that’s a completly different construction site

the main issue is that we can’t get the existing registries updated (they have a repository, I think xsf/registrar or so, but as mentionied earlier, no way to bring stuff to the website from there)

really? We use humans to verify the XEPs anyway because they advance to draft

even if the "tooling" is a ToDo list

because nobody is going to remember that there’s a registry for '30 nodes

or '50 nodes

it needs to be written down somewhere and people need to be aware that it exists and need to use it

even better if it’s in code

I am certain there is a way to update the website, we just need to poke the right people so that that part can be automated

larma, guess what

I’ve been doing that for the past six months

since one of my PRs got rejected with "that should be in a registry"

so, who is responsible then? board?

does it matter?

iteam doesn’t have the resources to do something about it

I haven’t brought it up with board officially, but I don’t think that’d be much use either

though maybe we should get board to pay MattJ to bring our infrastructure up-to-date

I heard we have some funds over

Kev, as an author, could you comment on https://mail.jabber.org/pipermail/standards/2020-February/037092.html ? 🙂

jonas’, I am a bit confused here: we do have an iteam, but they are not able to duplicate the automation that is already in place for xeps to the registry?

larma, yes, because something changed at docker hub

it would now require to grant docker hub privileges we don’t want to grant them

we can’t add another repository

so we have to use a different way to build the docker image with the static files

so how about we put the registry in the xeps repo as they will be updated only at the same times anyway?

😀

doesn’t help, we still need to set up a another repository on the docker hub side of things

(and connect that repository to github, which is the crucial step)

and even then, I’d not at all be happy with this "solution"

I don't use docker (for good reasons apparently) so I can hardly help there

docker hub and docker are only tangentially related

but they are

as much as I don’t like the container hype, that’s got nothing to do with it

docker hub is simply a crap CI infrastructure.

the web UI is also crap

(and who knows, the change may also be on the github permission model side of things, I don’t know)

and it happens to be the main/default repo for docker

sure

the image hosting is fine

just the automated building in *their* cloud is crappy

(I have no idea how they pay for the insane amounts of bandwidth this thing must eat either way)

"jonas’> but it wouldn’t catch cases like '84 which simply miss a submission to a registry", while it's still not perfect, isn't protoxep -> experimental (and other subsequent state change) supposed to help a bit(?)

Ah larma suggested that already

"jonas’> though maybe we should get board to pay MattJ to bring our infrastructure up-to-date" I'd be happy to do this, if he agrees. Or to find someone else

pep., '84 was changed in Draft, wasn’t it?

no wait, I’m confused

but '48 would be just the same actually.

yes human error is always a factor. It's never too late to fix it :) (once we get the registry back)

pep.: I *think* that paying MattJ would be potentially problematic because of renumeration and Board (I'd have to remind myself of our bylaws).

Kev: what if membership votes on that instead of just board?

Someone would need to check the rules about Board members receiving remuneration whichever way it comes about, I think.

Section 4.3 Compensation. Directors shall not receive any compensation for acting as such, but Directors shall be entitled to reasonable compensation for services rendered as an employee of the Corporation. The Corporation shall be entitled to purchase officers’ and directors’ liability insurance without violating these Bylaws.

There we go then.

Kev: I was thinking about this indeed

I'm happy if it's somebody else then

I'd also like to see iteam gain more resources, and if paying MattJ will give that, I'm all in

A couple of years ago we had an almost-offer of sponsorship in exchange for people helping the iteam with various tasks, but sadly it didn't come to anything.

Just to be clear, is the quoted text from the bylaws saying it would be ok, or not ok?

I read it as "ok"

me too

Yeah, you can't be paid because you,re a director but you can if you actually work in/for the organization

I'd like to have a clue about finances before though

I read it as the XSF is able to employ a Director in order to pay them for work they wouldn't do as a Director.

yupp

iteam isn’t director’s work, is it?

if it is, many a director is shirking their responsibility :D