XSF Discussion - 2018-11-19

the Observatory appears to be misbehaving. PHP errors on the front page. Jonas', would you mind kicking off that frontend docker hub project, to see if it goes away?

Guus, I can’t

it needs manual iteam intervention, it’s not pulled automatically from docker hub

uhoh

ah

I’ve tried to get iteam attention for weeks now about this :(

but in my mind, I already figured that we need to re-implement this on top of something less awful to maintain

What we see now is not a relatively simple front-end issue?

where's the code anyway? Docker pulls from a non-existing github repo?

https://bitbucket.org/horazont/xmppoke-frontend <-- that one?

err, bitbucket, not github

the xmppoke docker stuff itself on https://github.com/horazont/xmppoke-docker

the frontend lives here: https://github.com/horazont/xmppoke-frontend-docker

yes, but line 14? https://github.com/horazont/xmppoke-frontend-docker/blob/master/Dockerfile#L14

(I just posted the release note of Movim 0.14 on HN, for those that are interested https://news.ycombinator.com/item?id=18485822)

Guus, ha, nice

but why

I don’t have mit bitbucket credentials at hand to check

bitbucket never agreed with me

hopefully you can find out what happened later

Guus, I might forget, feel free to remind me later today (15:00Z+) or on wednesday

kk

why not put a static page up on xmpp.net, saying that we are looking for help maintaining it

Sounds good

I'd volunteer to just rewrite it. But I think someone else is already doing so

It seems to be broken.

I restarted both poke and the frontend, and the frontend refuses to come up.

Kev, that's probably because the docker instance is being composed partly from a repository that's currently unvailable.

before the restart, the frontend gave a different error.

Daniel: what makes you think that someone else is already doing so/

Someone told me

😏

Holger mentioned that Matt the jabber.at admin is working on it and also sent me a link that I haven't had time to check out yet

Guus: It was just starting it, rather than recreating the image.

https://github.com/mathiasertl/xmpp-test

daniel: But in the meantime, he said "right now, it's more or less on ice".

If that doesn't lead anywhere I might start my own. I want a test that then will result in a simple boolean to integrate it into the compliance tester

And if I write the tooling to create the boolean I already have what it takes to generate a more verbose output

I suspect that the frontend can't cope with being restarted without being recreated, or something.

Daniel / Holger: Nice!

Kev: Odd!

Until a viable replacement is available, it's probably best not to give up on our old instance though

It seems like someone's moved things around on the filesystem and broken things, but I don't remember who set this up.

I think I might have to grow the compliance tester into a more general thing anyway. During the sprint I wrote a thing that tests for cors headers on the HTTP upload

might take a while before the replacement is ready. Looking at the effort that has gone into the old, but also the new tings (100+ commits) ...

At some point it will make sense to integrate tls testers

Kev, might've been me?

Let's see if I can at least fix the backup script before I go nuclear on it.

As things stand the backup script is broken, and the README is wrong.

(And the container won't start)

Well, at least I've got the backup script working again.

Oh, the creation scripts only work if the instance is running. If it's not running, they'll refuse to install a new one.

Got it back to Warning: count(): Parameter must be an array or an object that implements Countable in /var/www/html/index.php on line 97 state, at least.

Thanks Kev

daniel, who is rewriting it?

the TLS tests can be easily based on testssl.sh, it has a JSON formatted output

that was the plan when I get around to do it, but I didn’t start

jonas' : https://github.com/mathiasertl/xmpp-test

Jonas' this is your 15:01Z callback ping 🙂

thanks!

apparently I changed my username and bitbucket doesn’t do redirects :)

jonas’: what I'm hearing is you broke it and it's all your fault

no

Lol <3

Guus, fix for the URL is pushed

thanks jonas' !

isn’t there a general race condition with multiple clients of the same account in a single MUC where when a MUC forwards a groupchat message to one of the clients and the client goes offline, it might get sent by the XMPP server to the other client ("most available resource")?

jonas' : both index.php as well as list.php don't handle a scenario well where the find_score query doesn't return rows. I'm assuming that's easily fixed. I'm unsure why no scores are returned though. To much work to be processed?

I don’t know

https://bitbucket.org/jssfr/xmppoke-frontend/src/84b48437dca24deaf7557c982c069015cf09a396/index.php <-- line 89 does appear to validate if scores is non-empty, but line 97 does not

I didn’t write the PHP code, and I don’t want to maintain it.

I'll have a stab at fixing the code to prevent the ugly php error

hargh, how do I fork in Bitbucket?

ah, by creating a new repo

oh, good, it uses mercurial.

not that hard when you just want to patch a thing

hg clone, do your thing, hg commit file1 file2 ...

sure, but that'll make me install hg.

I'm doing things in the text editor of bitbucket

Jonas', I think I just created two PRs against your repo, that _should_ fix the issue

Guus, can you try to try it for yourself?

the frontend thing contains a docker-compose file which you should be able to use

Guus, hm, I’m really not fond of maintaining this

I don’t feel comfortable in reviewing PHP changes, and yet I’m now somewhat of a gatekeeper

I'm to stupid for docker

somehow, my changes aren't picked up

did you re-build the image? with the correct tag?

probably not

ah, got it

jonas' : my commits fix php-related errors

likely not the underlying issue of the tests not being executed

but at least, it's failing more gracefully now.

mmm

jonas’ [15:20]: > isn’t there a general race condition with multiple clients of the same account in a single MUC where when a MUC forwards a groupchat message to one of the clients and the client goes offline, it might get sent by the XMPP server to the other client ("most available resource")?

no, there isn’t, I figured it out by now

due to: 8.5.3.2.1. Message For a message stanza of type "normal", "groupchat", or "headline", the server MUST either (a) silently ignore the stanza or (b) return an error stanza to the sender, which SHOULD be <service- unavailable/>.

sorry for the noise

Groupchat messages are not redirected to other resources

Sorry, great :)

Also "silently ignore the stanza"

Silently ignore that option

Implementation defined behavior.

IB!

at least not UB!

Silently ignore buggy implementations

MattJ: they are not buggy if they are compliant

You think so? :)

MattJ: by your definition, mod_carbons and mod_mam are buggy as well

They are

And the XEPs are buggy too. And the RFC.

TCP/IP is buggy, but (mostly) follows spec

Abandon hope

Humans are buggy

it’s all buggy

Not all

The stuff we haven't written yet is perfect

It's buggy all the way down. A fractal of bad protocol design

Ge0rG: great name for a client!

MattJ: "Buggy"? 😀

Maybe you should just start from scratch, call it matrix or something

Bug is a kind of dance?

I'm sure you won't hit the same problems :)

> have you tried the Buggy messenger? All of them!

-y as suffix is hip again?

hippy

Buggy, the zimpy messenger.

My phone auto corrected zimpy to "gimpy"

when was the last XEP that made it to DRAFT, just out of interest?

commit c58d3bbf5ed18e95b374e3bc1db9a85de40f2661 Author: Jonas Wielicki <j.wielicki@sotecware.net> Date: Thu Jan 25 08:54:39 2018 +0100 XEP-0387: Move to draft

that’s a good candidate

1806, it was a cold winter that year

ha, the compliance suites

lol :D

ok and now a real xep

please

must have been in the dark ages

I’d h ave to do more magic than that, but it might be May 2017

or March?

march

back when Sam was still editoring much -- it’s not my fault, I swear!!

Getting compliance Suite into draft was hard enough!

yes, XEP-0233 was the last one moving to draft before that, that was in Mar 2017

Speaking of compliance suites, someone should do some for this upcoming year (even if they're basically the same) so that they don't appear out of date

yeah, I’ve got that on my todo

If it's the same, can we just update the year?

next council will decide

(but I’m in favour)

I think I was missing one XEP when explaining what's relevant for mobile to haeckse last weekend

Oh yes, it's missing 0184

I don't even understand how we could miss that.