XSF Discussion - 2018-05-10

pep.: xmpp:gpn18@conference.conversations.im?join

Maybe Link Mauve is interested as well? Are you going to gpn?

daniel, I was planning to, but ended up not planning anything. :x

(surprinsingly enough)

ELOGIC

Man (in club): "Hey DJ, play Love to Hate You" DJ: "No" Man: "I demand it under my Right to Erasure"

m

I suppose erasure is a band or something?

I’m not sure if I should be happy that I can infer that or whether that’s a sign that my mind is twisted.

fippo, You about? I've some XEP-0220 questions. (Mostly about errors).

Anyone else might know, too. I've an amusing... Bug? in Metre. Sequence is roughly the following:

Metre -> Other : Outbound connection, bidi-enabled, sends dialback request.

Other -> Metre : Inbound connection, SASL-authenticated, bidi-enabled, sends dialback verify request.

However, Metre has already closed its outbound, seeing a usable and authenticated bidi-enabled inbound session. So it rejects the db:verify because it's referencing an id it knows to be closed.

So, questions:

1) Should a db:verify that's valid, but referencing a closed session, be returned "valid", "invalid", or "error"? And if "error", what error code?

In fact, that's kind of it. I may well tell Metre to leave any session that's even requested auth up when it sees a bidi possibility, even if it feels less optimal, but I'm curious about how to handle it nonetheless.

dwd: why does the fact that the connection associated with the id influence the validity of the dialback key?

fippo, Paranoia.

fippo, Metre won't respond "valid" unless the outgoing stream exists and is in the right auth state (REQUESTED).

Dave Cridland, btw, is Metre multi-threaded, at least for TLS handshakes?

Now that my personal server got two cores, I was thinking about using something in front of Prosody for that.

hrm. in that case you might need to keep the connection around for a bit longer. or hope that people like me stop using old certificates :-)

Link Mauve, Nope, not yet.

Ok.

Link Mauve, Of course, it'd offload the TLS from Prosody, an you can run Prosody as a component which'd possibly work neater.

Link Mauve: how much CPU time are you currently spending on TLS handshakes?

Mind you, you can also run Prosody as a component host for another Prosody acting as the user, so...

Ge0rG, enough that s2s doesn't reliably establish before its timeout on server restart.

It is a real toaster!

Link Mauve, I do have a bug there - although Metre uses libunbound in async mode, it seems to have a limit in the number of queries it allows in-flight (or maybe it's a sync interface hidden in the other thread).

Hmm...

Link Mauve: send pictures!

Link Mauve, I'll get around to fixing that one though.

Ge0rG, I'm not at home right now, maybe later!

Link Mauve: prosody might not like it bit handling TLS anymore? What's the thing again in sasl that needs stuff from the TLS session

Not* handling

pep., SCRAM-SHA-*-PLUS, but if you can transmit that it's fine.

Metre doesn't do c2s tho

Or does it?

In this case Prosody would be the one doing it.

Metre would only establish the connection.

AFAIK.

Metre won't even listen to C2S ports.

Metre just does S2S and '114. (Well, and '368 and '361)

Oh.

So yeah pep., your question doesn't make sense, Prosody would still listen there.

Ok

vanitasvitae, "a function that teenagers find neat" - ah, the youth of today, eh?

Any news on fixing Travis?

vanitasvitae: I actually use signal only because it has this feature. Thanks for the compliment!

remko, you're welcome ;)

vanitasvitae: I find it useful to e.g. send passwords to someone. I know they won't delete it themselves, so i do some extra security management for them :)

I'd not really call that security management but I get the point

It is if you trust them to be incompetent rather than malicious.

e.g. relatives

exactly

although worded differently when in presence of

If you say so :)

Sometimes, honesty is best

as usual, whenever I step in to a thread on the list, it derails quickly.

Moth to a flame.

yeah, i'm still going with the assumption that i'm the moth, so it's not really my fault.

Steve Kille, hey, your XEP-0369 update makes non-editorial changes. I would like to keep the version numbers semantic, so would it be okay if I change the number of that release to 0.10.0 instead of 0.9.7?

(The idea is that changes to z in x.y.z should only be of editorial nature, to make it easier to guage the impact of an update)

jonasw: that is fine. I was not sure. I had the sense that small technical changes should not bump the middle number, but happy to follow best practice

Although there was quite a bit of text change, there was nothing radical changed

Steve Kille, cool, will do then

Smallest number is editorial changes only. Largest number is Advancement. Anything else is middle number.

got it

Is XEP-0404 still available? I've got a non-April 1st use for it that I want.

Kev, yes

jonasw: Could you 'reserve' that one for me please?

Kev, yeah

I'll explain in a bit, just on a call at the moment and thought of it.

no worries

next one is going to be 403 anyways

and since there’s no other protoxep in the queue...

There is a problem of certificate for jabber.org, we must create a ticket?

https://github.com/stpeter/jabberdotorg/issues/2