XSF Discussion - 2020-03-30

re unicode version fun: https://labs.spotify.com/2013/06/18/creative-usernames/

Nice

Huh, what's that random draft they link to?

I thought nodeprep et all was some RFC somewhere

Oh, 2013

Where did the RFCs go.....

Sigh

Which?

where they belong, on ietf.org? ;)

https://xmpp.org/rfcs/rfc6120.html ‽

Were they in yet another container?

No, looks like they were on the host filesystem

NOOOO! The only good rendering of our RFCs is gone!

Restoring now

MattJ: do you have access to the host again?

And they're back

!praise MattJ

!praise backups

nooo the worst rendering of an RFC ever is back ;-P

MattJ: Can you also pull the docker containers for xeps and xmpp.org plz

Yes, I'll do that now :)

ta

I assume this was previously just in cron or something

very likely

Would Someone™️ like to review, merge, maybe even test, the wiki PRs I made?

given the insane amount of pulls the xeps docker image has

pep., done

Any opinions on what an entity should return if someone includes an unrecognised form field in a form submission?

> When receiving a search form, the Search Service MUST ignore fields with a var value it does not understand. XEP-0433

Either a) ignore it or b) an error

if that’s not an option then I’d suggest feature-not-implemented with an application-specific error condition.

Yes, I was going with (b) an error

not-acceptable?

Just wasn't sure what error

or what jonas’ said

or bad-request

XEP-0004 says not-acceptable if data validation fails

What about if you don't include an optional <field/>?

feature-not-implemented, not-acceptable and bad-request were my shortlist

MattJ, roll a dice

Do 22-sided dice exist?

on every reply :)

6-sided and divide by 2?

jonas’, I was thinking about 433, could it use order-by 413

pep., the wire format?

yes

maybe

though that’s not in a data form

so it doesn’t really play well with that

Maybe that's a point against data forms :x

(query (form, order-by))

Zash, so you can’t configure order-by in the form then

MattJ: But isn't that for when you know about the feature, but you don't implement it?

pep., I’d be fine with ditching data forms altogether

... whyever you would ever do that

jonas’, doing what MAM did backwards? :)

simple text nodes → FORMS!

why did MAM do that?

Zash, it's when you know about the namespace, but don't recognise the element (according to the RFC)

jonas’, summit consensus iirc

why?

XMPP needs ADRs

we really really need Design Considerations sections in *all* the documents.

We really need to record the complete brain states of everyone involved

https://xmpp.org/extensions/xep-0313.xml#revision-history-v0.3 Oh no, I did this

I AM SO SORRY

Haha

So, Summit 2014 minutes?

hahaha

Why are some events by year but Summit is numbered?

who knows..

CCC is also numbered

Eh do we even have minutes from then?

Helps seeing how old it is :p

https://wiki.xmpp.org/web/Conferences/Summit_15

That was the year of binary sponsorship

https://wiki.xmpp.org/web/Summit_15_Minutes#mam_search

jonas’, if there's a "why" recorded, it'd be there ↑

I suspect it was probably just to avoid reinventing the wheel

A query is a bunch of data slots, the client needs to discover what slots the server supports

Each slot has a type

Sounds very much like forms to me

or schema

disco#info features

The XML bikeshed should be ${shuf -n colors.txt}

Forms are nice for ad-hoc stuff that might be shown to users

And you still had to put RSM outside of the forms

I don't know anything that lets users type things into MAM search yet

Because it doesn't play well

(whatever the reason)

RSM being a thing seems logical to me

https://xmpp.org/extensions/xep-0413.html#MAM heh, does this even work

implemented anywhere?

Ah it's with pubsub

fear

Now is the best time to do a radical cut.

Magic per-device offline messages when?

I gave up on that tactic

The magic is too magical without a way to identify devices, which means clients would have to update to bind2/ng to take advantage of it - at which point they should just implement MAM and gain control over what they receive

Apparently the resource is now the device ID and things break if the server uses its spec-given right to set it to whatever it wants. :(

What things?

I ran into a bug where Dino doesn't deduplicate outgoing messages correctly because the messages from the archive has the server-issued resource but the local archive has the requested resource.

obviously dinos fault

+1, file a bug and toggle Prosody's default to rewrite resources

:>

I've threatened it before and I'm semi-serious

This one weird trick. Client devs hate it! :D

Users are actually the main complainers in my experience (they want their resource to be Home, Work, or whatever)

A simple device name spec would solve that

Resources are session identifiers only

We have a device name spec

XEP-0030

File bugs :)

So PAM (Pubsub Archive Management) is taken...

I'd like to have resources use something like this by default: https://github.com/moby/moby/blob/master/pkg/namesgenerator/names-generator.go

human readable, easy to distinguish, not leaking more information than necessary

Resources shouldn't be visible to users

yeah that's not for users

MattJ, they are visible to admins reading logs though

for server admins to debug

Inb4 resource := userpart + serverpart

We're not doing that.

Objections to adding an 'id' field in MAM to fetch a single message? or should that be a separate iq?

Combining 'id' with any of the other options is likely useless

Seems reasonable to me.

mam:3 incoming ? 😱

Avoiding bumping the namespace, but yeah

So get your requests in now before it's too late :)

Please do post on the list for feedback maybe before merging the PR :)

why, it’s Experimental :>

;)

To avoid having to publish yet another revision.. not that revisions are expensive, but the effort of doing the PR is expensive

Once you're in.. you're in

might as well group related efforts

Much evidence shows that revisions are expensive

increasing a number?

Doing the work

What I said, then :)

Interesting: https://github.com/mar-v-in/libomemo-c larma

jonas’, does openssl even support EXPORT ciphers anymore?

if you build it correctly, I think so

(is "correctly" the right term?)

I think so. appropriatly?

(this was a allusion to the fact that a libssl with EXPORT ciphers is most certainly built incorrectly nowadays)

Zash, you could use the libssl from here: https://testssl.sh/

oooh