XSF Discussion - 2020-06-14

Guus, ralphm, Seve, pep., MattJ: I need a quick opinion poll from Board. I figured out that GitLab.com may be very helpful in solving technical issues we’re having with the editor process. I made a quick proof of concept in a private space on gitlab.com and it’s looking pretty good. We’d like to experiment with that over the coming days and eventually, Board would probably have to agree to create an official XSF resource on gitlab.com. I don’t want you guys to feel like I’m doing a "fait accompli" style of thing so I want to let you informally know in advance and stop me from doing it (MattJ is on-board "if it solves issues", iteam-wise).

(as a sidenote, as a leftover of the rebellious energy from the whole github.com-taken-over-by-microsoft-deal, I still own the gitlab.com/xsf namespace, to which I also gave MattJ access last night, so we actually have a good resource there available the XSF could take over.)

Always +1 on experimenting. Curious what problems using GitLab solves that GitHub doesn't.

ralphm, the CI pipeline features of GitLab roughly combine what we can do with travis and docker hub in one thing, but without the extensive permissions we need to give dockerhub to add more repositories

(the latter is probably the fault of the terribly coarse permission model of github)

and it can do even more beyond that

like rendering HTML previews of XEPs which were changed in a merge/pull request and attaching them to the webinterface so that you can take a look at the result (and people don’t have to self-host their rendered documents anymore)

I've used GitLab before and know you can do these things, but I am interested in understanding why this would be preferable over similar features in GitHub.

there are no similar features in GitHub, or if there are, they are less convenient

GitHub started to have a docker image registry / artifact store, but you need to be authenticated to access it (even if the repository is public)

it’s all over all much less flexible than GitLab in this regard; they have much of a headstart on this

for stuff like the rendered HTML previews, we’d have to write custom bots and integrations, while gitlab can simply expose rendered data from a CI job in a MR in the web interface

and, most importantly, we can do incremental builds with the GitLab tooling, by using the cache mechanisms. This cuts down the build time from >1h to a few minutes.

Sounds like something at the very least worth exploring. Assuming things are public, it might be good to think about needing to earmark artifacts as non-normative / experimental during testing, to avoid having two competing sources of truth.

Guus, indeed

That's helpful, thanks!

I said in another channel already I was happy with the idea :)

Thanks for the feedback all. We’ll then proceed with experimentation.

Second all of the above!

jcbrand, sorry, only seeing your message from two days ago now: you can configure matomo to use server logs instead of a client side script. No cookies and also tracks people that have adblockers. You still need to mention it in the privacy policy, but don't need to bother users to confirm it. See https://matomo.org/log-analytics/

jonas’, experiments are fine, but https://xmpp.org/extensions/xep-0390.html#revision-history-v0.4.1 just confused me, so maybe not deploy experiments to live?

what

larma, well spotted

I’ll clear the runner caches now

the downside of incremental builds, eh

that and a "Random change." hidden in another XEP are the only glitches, I promise :)

(and they’ll vanish with the next clean rebuild, which takes about 45 min due to the PDfs)

Yeah, was just confused for a short time when the email said 0.3.1 and the XEP said 0.4.1 is latest version. But the description of the change cleared things up 😉

the email itself was also an artifact of the caching

yeah, but it remembered me I had pending comments on that XEP 😀

:D

larma, please keep them for a moment

0.3.2 is going to be published later today

(it’s already queued in the pipelines)

(though that’s only a patch-level change and won’t emit an email... and it’s really minor)

larma, so probably better to just start a new thread on standards@ with your comments, I appreciate them :)

are editors payed by the XSF?

nope

I don't think anybody is payed by the XSF atm

iteam has a budget, but it goes to paying the servers, not the people

I mean for a recurring task

Recurring tasks? Automate them!

Still nobody is payed to automate them either

*paid also

oops

Will there be some special meeting for the gitlab experimemtations?

no

and whn on 17th is the decision taken? Just over the day spreaded?

*when

eevvoor, does it matter for you?

just because editors think it’s a good idea will not mean that any switch is hgappening

board needs to ack it first

*hg*appening?

when the switch happens, an announcement will be made before

Zash, :P

+1 for the change fwiw

Personally I don't care what editors use as long as I'm not affected or required to have some account with some 3rd party to contribute.

eevvoor, it's just the board meeting

Zash, define third party. GitLab is kind of a third party, isn’t it?

(just like github)

jonas’: unsure if this is helpful at this stage, but did you consider non-gratis solutions to address the issues you've been working on?

jonas’: Read: As long as I can email you patches I'm reasonably happy.

Zash, right

Guus, did not see the need yet, since the free tier on gitlab.com is sufficient for what we’re doing

it might be useful to pay or host our own CI runners at some point if the shared ones are crappy. at least with gitlab we have that opiton.

XEP-????: CI dispatch over XMPP .. when? :)

Kk. Just wondering if Sam's concerns could be alleviated by using some kind of paid alternative. I currently don't have formed a preference myself at all, just wanted to point that solutions as afar as I am concerned need not be gratis.

Guus, GitHub is not going to become better by throwing (insignificant) amounts of money at it. Sam’s main concern is moving off GitHub.

To be honest, I think there’s a major difference in quality of the interface itself, beyond the CI integration, which would make GitLab favourable for our use case

Argument based on popularity?

yeah..

GitLab is a DevOps platform, GitHub is a software development platform.

Which I personally like to dismiss

The editor work is much closer to devops than to software development, and it shows.

github is trying to get closer with github actions (and if you’ve ever used the Azure DevOps platform, you know where they got their inspiration from), but they’ve still got a long way to go

I agree

There's some merit in not changing venues to avoid confusion etc. I that it's more important that we can improve the performance of our processes though. ✏

the registries have been a pain point and they still are and we need to fix them

we might be able to do that by resorting to gitlab CI on github

Just my 2ct: If that step would be made, we may think about unification and move the Wiki and Trello to it as well. So we have everything in one platform. I think Gitlab has those functionalities as well.

the value which would be added by migrating to gitlab for the repositories themselves might be nice though, too

emus, it does indeed have issue boards and wiki functionality

Yes, saw that at KDE

and there would be some merit being able to cross-reference issues, PRs and agenda of board/council

I'm not a fan for changing thing just to consolidate our on one platform.

sure

The Venue is the mailing lists.

that, too

I dont see any interest in staying attached to multiple platforms either if it's just not to have a single platform

I'm personally open to experiment with wiki and all

I’m not convinced that any non-wiki.xmpp.org location would be a good replacement for our wiki. mediawiki is a lot richer than either GitLab or GitHubs wiki

and we’re using that richness

I also wonder if we should maybe do away with the PDF versions of the XEPs and replace them with epub or something.

A bit related offtopic: Therewere claims about drafting the newsletter on Github and not in the wiki anymore (excluding non-tech people, what not true in my view). That would make things easier and also may offer to host translations by ourselfse and also reduce formating from one platform to another. But I see thats not the point right now.

Building them has been a pita quite a while now

pep., look at the main page

jonas’: are we using specific mediawiki features? I'm genuinely curious ✏

I mean, you can probably dump html into gitlab/github markdown to some extent, but that would kind of defeat the point.

the separation thing?

(and I guess it’d still filter much of that away)

colours, the layout, in general, the structure

markdown is an html superset right? :p

yes, but people tend to sanitise it afterwards

now I need to test if you can use style="" in gitlab wikis

how easy is it for non-technical users to contribute?

probably easier because the mediawiki markup is confusing and strange

does gitlab have a thing that lets you edit files on the web?

yes

yep

Zash: newsletter? Make a github account what most people here have anyway. create a comment or file change. I also have a simple tutorial if someone needs it ✏

If making a github account is required then I'm opposed

it is

everything in the visual interface. You can as always just drop your meassge in the muc

though you can also send patches to commteam@ ;)

Zash: Making a wiki account is also necessary and needs permission, so Github is actually more general and open

jonas’: You can, but we prefer people to just make it from their side with more project insights and intrinsic motivation ✏

emus: a wiki account is attached to the XSF, not an unrelated 3rd party corporation with a ToS

(instead, it is related to the XSF. does the XSF have a privacy policy for the wiki?)

Zash: thats true

There have been actually only one claim on this

most were very happy with the change

jonas’: don't think so. that was discussed quickly some time ago in a board meeting, somebody (TM) needs to do something about it

arguably, from an objective observer, a clear ToS may be preferable over complete lack of tos ;)

indeed

Zash: as said you can always just comment in the MUC

Anyway, we are getting away from the Gitlab discussion

jonas’, sure. but still, first party maybe(tos) + 3rd party tos > first party maybe(tos)

I’m not sure if `>` means "better than" in your case

more ToS-es

In Gitlab would be our ToS? or Gitlab.com?

gitlab.com

the XSF could add stuff like CoC on top if it wanted (stuff that would be valid for all XSF managed spaces say)

reminds me that the clabot feature would have to be reimplemented in CI on gitlab.com, since there’s no such thing for gitlab for some reason

isn't that "just" another service to plug to webhooks?

not quite, since it needs to communicate back to the service + do some auth stuff

larma, ghost-revisions of '390 are gone

when we are considering to move to gitlab, why not to our own instance of it? solves the issue of needing a third party 😉

larma, resources

don’t want to burden iteam with more stuff

Ah, right. Yeah, iteam is probably overworked anyway (at least iteams tend to be)

fairly certain that’s also the case for our iteam ;)

I think it is very difficult to have main Gitlab for XSF, because a long time ago there was a JIRA and we know where is it...

I think that the current step must be to solve all current issues...

Missing documents...

no

it most certainly does not have to solve all issues

then we’ll never find a step which does that

It is a first point to show XMPP Standards Foundation (and XMPP) quality, reputation, etc. - https://github.com/xsf/xmpp.org/issues/created_by/Neustradamus

Why some companies use Matrix now? :/

A good website is the showcase...

Not up-to-date / complete are bad points.

We should simply fix everything and become better than Matrix in all regards. This should happen today.

More feedback on social media

something something mobile clients

More Matrix in XMPP. Apart from the fun, the critics are legitamite in many points.

>More Matrix in XMPP. What do you mean?

The money/marketing part? When I tried Riot once it wasn't really convincing me and what I've heard about running a Matrix server sounds like this is horrible.

Was just stupid fun in relation to the previous messages