XSF Discussion - 2020-05-15

6121 doesn't seem to include a rationale as to why it's not possible to add multiple roster items at the same time. Any idea?

Slightly more fine-grained errors? Even if it's already possible for one error reply to mean multiple things

Hello everyone, I'm Aditya Borikar, a candidate for gsoc 2020. I plan on having my weekly logs put up at https://adiaholic.github.io/gsoc2020/ Is it possible that we add this to planet.jabber.org ?

ralphm: ^

adiaholic_: in case you don't get an immediate response from ralphm in here, you might also write directly to ralphm@ik.nu (Jabber and e-mail).

Ge0rG: Alright. Thanks alot 😀

adiaholic_, I added the feed and it should pop up for the next poll cycle. Thanks!

🎉️

Yayy !

MattJ, I only ask here because we talked about it in another channel (I don't have time to allocate right now to this), will it be possible in MAM to configure storage duration etc.? So it's not just a boolean (store / don't store) ✏

Is that at least something you'd see in or around MAM

pep., yes, but not in 313

store/don't store is out of 313 now - I don't like the routing/reliability implications those preferences have, and most clients (probably rightly) don't expose them to the user

Message retention however is an important preference that should be available to the user

I almost added it to 313 in the latest revision, but it got too complex and I left it out

I'm planning an account preferences XEP, so I'll probably base it on that

Why xep-0414 (crypto funcs) says BLAKE2b256 is MUST and BLAKE2b512 is SHOULD while implementations do the opposite? openssl for example doesn't have BLAKE2b256 at all.

AIUI, blake2b doesn’t have specific bit lengths, it operates at some length and then the application truncates the final hash to the length of its choice.

Link Mauve: I checked just openssl in details. There is hardcoded digest size to 64 bytes for blake2b

That sounds unwanted.

reference implementation supports variable digest length but it's terrible slow from my tests

But I don’t know much more about this hash function, I just use the RustCrypto version of it.

rion, I’m fairly certain to have tested that with python, which AIUI just uses openssl under the hood for the more "fun" hashes

rion, but I suppose we can change that if it poses an issue in practice

I'm only aware of blake2s256 and blake2b512 in OpenSSL, and I have had no reason to use them yet so that branche isn't merged

Zash: yep. for 256 there is blake2s256. I believe it generates a different digest in comparison to blake2b256

Well I would agree to make blake2b512 MUST instead. I guess it should still work on x32 systems with reference implementation. Just not optimal. But it's quite widespread.

rion, can you make a PR against the XEP, ideally with data about various implementations and their support levels?

e.g. what about libressl?

problem solved but why did you change the port from 5269 to 9269?

xnamed^, it's been like that for as long as I can remember

jonas’: not supported at all by libressl

jonas’: I'll make a PR as soon as I finished with QCA which touches many implementations

thanks

Is someone around who could give me admin on wiki.xmpp.org?

MattJ, MattJ should be able to ^

(sorry, scnr)

Yes, I don't feel like poking around in databases, but I guess it's a last resort

MattJ: are you talking of wiki admin or ssh on the box?

Wiki

My account got lost during the Great Crash, and when it got recreated it was without admin privs

MattJ: according to https://wiki.xmpp.org/web/index.php?title=Special:ListUsers&group=bureaucrat you need to ping Alex, Kev or Xsfadminaccount

.

Don't worry, I just opened the docker stuff for an unrelated reason and found the password for xsfadminaccount :)

(which I didn't know was a thing)

Oh, I added you to admins as soon as you asked, pretty much.

Ok, thanks

Well, as soon as Ge0rG replied, actually.

A nice example for the effectiveness of name-dropping.

I'd rather you hadn't, TBH. I'm off work at the moment for personal reasons.

But yes, it was effective.

Kev: sorry. Have a nice day :)

I turned xmpp.net back on, on the old server

Don't spoil everything

inb4 load over 9000 as usual

seems to work

Yay

xmpp.net is back!