IRC office hours/Office hours 2015-03-23

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

Log for Flow[edit]

[19:29:58] <quiddity> Greetings all.
[19:30:12] <quiddity> Welcome to the Office Hour for discussing Flow.
[19:30:41] <quiddity> The main focus of this discussion is intended to be the upcoming conversion of LQT to Flow, on
[19:31:00] <quiddity> See announcement and links at
[19:31:31] <quiddity> But the team is also ready and happy to discuss other aspects of Flow. (Main page at )
[19:32:17] <dannyh> hi, I'm Danny -- product manager for the Collaboration team at WMF, working on Flow
[19:32:32] <superm401> I'm Matt, software engineer.
[19:32:35] <dannyh> and I'm happy to talk with you, answer questions etc.
[19:33:06] <quiddity> Does anyone have any questions, to get us started? :)
[19:33:39] <superm401> quiddity, you want to start with an overview of the conversion?
[19:34:15] <quiddity> Sure. The plan is currently drafted in more detail at
[19:35:16] <quiddity> The team intends to begin the conversion on April 6 (Monday), with the assumption that all blockers are fixed before then.
[19:37:29] <quiddity> They'll begin converting just 2 pages (mw:VisualEditor/Feedback and mw:User talk:superm401 (the volunteer-account usertalkpage of one of our engineers)
[19:39:38] <superm401> Task tracking the overall conversion is
[19:43:24] <YuviPanda> quiddity: would you rather fight 100 SULF sized ducks or one horse sized Flow?
[19:43:52] <quiddity> Ha! One horse-sized Flow.
[19:44:09] * YuviPanda notes down, will use against quiddity at his RfA.
[19:44:12] <quiddity> I do not envy keegan's mailbox.
[19:44:30] <YuviPanda> Do not covet thy neighbor’s mailbox
[19:44:32] <dannyh> I would sit the ducks down and encourage them to work together as a team
[19:44:34] <Keegan> Grind grind grind grind
[19:45:03] <dannyh> -- or -- I would romance them
[19:45:04] <quiddity> Although, a horse-sized project like Flow, is somewhat made up of 100(0000000) duck sized issues...
[19:45:11] <thedj> so can i ask what the point is ? Is it just to get rid of LQT ?
[19:45:38] <dannyh> yeah, there's two points, and that's definitely one of them
[19:45:59] <dannyh> we're going to have to decommission LQT at some point
[19:46:09] <Negative24> I see it as a big step in Flow development
[19:46:09] <thedj> has there been an increase about complaints on LQT, or more maintenance than desired ?
[19:46:13] <dannyh> and we might as well start doing it now
[19:47:08] <dannyh> honestly, it's just silly for us to have 3+ different ways for people to talk on the same wiki
[19:47:38] <dannyh> so it's partly to get started on that process
[19:47:51] <dannyh> and then the other point is to have another place where people are using Flow
[19:48:13] <thedj> ok, do we have metrics to show that Flow can be as effective as LQT currently is ?
[19:48:17] <dannyh> which gives us real use cases where things are/aren't working well
[19:49:43] <thedj> my point being, are we aware of if the user is going to experience this as being 'upgraded' or being 'downgraded with betaware' :)
[19:49:46] <dannyh> mediawiki is going to be the first really active wiki where we're doing LQT -> Flow conversion
[19:50:09] <dannyh> my guess is that it's going to be seen as a lateral move at the moment
[19:50:14] <thedj> ok, so we don't know and are going to heavily adjust where required during the process ?
[19:50:26] <dannyh> "using betaware" --> "using different betaware"
[19:50:46] <dannyh> and then the different betaware is actively being worked on and improved
[19:50:56] <marktraceur> dannyh: Until it isn't.
[19:51:02] <thedj> ---> is a learning curve here, and bad + learning curve == terrible :)
[19:51:13] <superm401> thedj, yeah, we've been very open to feedback in general.
[19:51:16] <wctaiwan> thedj: it's kind of irrelevant, though. LQT isn't being developed. Having developers use Flow will help / force them to fix whatever is wrong with it. is probably best-equipped to handle any rough edges.
[19:51:33] <superm401> For example, we were originally not going to do a TOC, but we looked into it further in response to feedback, and decided to.
[19:52:27] <thedj> ok, so what 5 things are we the most worries about/looking out for ? (assuming the conversion script works)
[19:52:35] <thedj> worried.
[19:52:46] <dannyh> the conversion script is pretty solid
[19:52:47] <wctaiwan> I guess you could argue that we should wait until we're absolutely certain Flow is at least as good as LQT, but I think such a comparison is difficult. LQT has the advantage of familiarity, and people are more willing to accept its faults because people are used to it.
[19:53:09] <dannyh> we did our first conversion on Officewiki, and then we've done a lot of test conversions for Translatewiki and
[19:53:47] <superm401> marktraceur, part of the LQT conversion is actually committing to not leaving unmaintained code out there. Although LQT did not fully work out, we're making sure that data is not abandoned in an unmaintained product.
[19:53:54] <dannyh> the one thing that is unknown right now is being able to convert all existing LQT notifications into Flow notifications
[19:54:12] <dannyh> the watched pages will all copy over, so you'll still be watching the same pages/threads
[19:54:30] <dannyh> but this will be the first time we're converting existing notifications. it'll be interesting
[19:54:42] <superm401> It's possible we may have some further Parsoid issues (obscure content it is not yet able to handle), but we have plans to deal with it if that occurs.
[19:54:52] <superm401> We've dealt with all the issues like that which happened during testing.
[19:55:10] <quiddity> (test conversions can be seen at vs )
[19:55:11] <thedj> one of my biggest concens right now is spam... About 9/30 notifications in my list is spam from the hovercards talk page...
[19:55:40] <thedj> which is a lot of noise for 1 page out of the dozens that i'm watching.
[19:55:50] <jorm> ugh, i had to unwatch that.
[19:56:35] <thedj> and this probably isn't easily solvable right ? because the notification system wasn't ever really designed to handle that ?
[19:56:37] <quiddity> thedj, I guess you're asking for a way to: watchlist a page, without *also* receiving (certain types of) Echo notification for it?
[19:56:57] <quiddity> i.e. a more powerful set of watchlist/notification configurations.
[19:56:58] <dannyh> we're going to be working in April on a more sophisticated watchlist/notifications system, to give people more options about how they're getting notified about new topics on a page they're watching
[19:57:00] <thedj> quiddity: no, i'm asking for a way for stuff that gets deleted to be removed from my list of notifications.
[19:57:01] <superm401> thedj, yes, it's solvable if this is a general issue people are having. We've already adjusted the Flow notifications.
[19:57:17] <Negative24> How is this conversion going to look in watchlists and will their be email notifications about changes?
[19:57:59] <thedj> hovercards gets about 2 spam messages a day right now, and even though the flow messages are deleted, the 'new topic' notifs remain in my notifications.
[19:58:00] <quiddity> thedj, ah, I'll file a bug for that.
[19:58:14] <quiddity> (if there isn't one already)
[19:58:36] <dannyh> yeah, thank you for mentioning that -- I'm not watching Hovercards & I haven't seen a lot of deleted spam so far
[19:58:49] <dannyh> that is annoying and ought to be fixed
[19:58:52] <ebernhardson> is that also a problem where the spam filters need to be adjusted, or flow needs to support more mw spam prevention tehcniques?
[19:58:56] <superm401> I don't believe it will show in your watchlist, right ebernhardson?
[19:59:00] <superm401> I know it will not in RC.
[19:59:16] <superm401> Because they are old messages, just changing systems.
[19:59:19] <ebernhardson> superm401: all the new pages will have old timestamps, so nothing about the conversion should show up inthe watchlist specifically
[19:59:26] <ebernhardson> s/pages/topics/
[19:59:47] <superm401> Nor will there be an automated email notification for the conversion itself.
[19:59:56] <ebernhardson> we should double check re-notificaions though, i'm not sure if we have ensured those wont be sent out
[20:00:18] <dannyh> Negative24 when there's a new post on a conversation that you're following, it'll show up in your watchlist -- and new topics starting on a page that you're watching generate Echo notifications.
[20:00:34] <dannyh> Negative24 you can also get/turn off email notifications, it's in Preferences
[20:00:41] <superm401> dannyh, but that doesn't include new posts from the conversion itself.
[20:00:46] <superm401> Or at least shouldn't.
[20:01:13] <dannyh> superm401 right -- it'll just be new posts
[20:01:24] <thedj> abusefilter and captcha work ?
[20:01:26] <superm401> Right, normal posts made after the conversion is done will trigger notifications.
[20:01:29] <ebernhardson> thedj: yes
[20:01:40] <thedj> ok. that should help with fighting spam
[20:01:52] <superm401> thedj, yes, except that you can not tag posts that are allowed through. You can just pass/reject.
[20:01:54] <ebernhardson> thedj: flow currently cupports abuse filter, confirm edit, spamblacklist, spamfilter and spamregex
[20:02:19] <thedj> ebernhardson: we need to make sure that's in the faq/announcement/docs whatever
[20:02:24] <quiddity> (details at )
[20:02:30] <thedj> cool !
[20:06:21] <thedj> one thing that still has me worried a bit is inpage and inarchive search...
[20:06:33] <thedj> this still seems very weak in the current version of Flow
[20:06:34] <superm401> thedj, in-board search is our top search priority.
[20:07:02] <superm401> This will include both stuff that's already visible on the page, and farther down (i.e. stuff behind infinite scroll will be brought in if it matches your search results).
[20:07:10] <ebernhardson> it has been progressing, but a bit slowly because there are no engineers assigned the the foundation search infrastructure at the moment. We are working on it though and getting closer every week
[20:07:13] <superm401> We've already implemented the back-end code for this.
[20:07:55] <superm401> The initial back-end code is pretty much done though, just needs code review (and any tweaks that brings up).
[20:08:45] <thedj> are we waiting for that or going forward in parallel ?
[20:09:35] <superm401> Waiting
[20:10:12] <dannyh> yeah, unfortunately we've been held up by it
[20:12:48] <Negative24> How will automated posts by bots and mass message be handled in Flow?
[20:13:18] <superm401> MassMessage already supports it.
[20:13:40] <superm401> Bots will need to be adapted. This is already in progress, but will still need a significant amount of work. We are supporting the bot writers and library writers on this.
[20:14:05] <legoktm> o/
[20:15:15] <thedj> hmm, i notice some of the menu items have no hover state at all.. that's slighly confusing. filing a ticket on that.
[20:15:49] <quiddity> I recently checked through the 1st page of contributions of accounts marked as and couldn't see any usertalkpage edits (at LQT boards).
[20:16:46] <Negative24> Just LQT boards?
[20:17:27] <quiddity> That's all that is being converted, in the near future, at
[20:17:54] <legoktm> I'm showing up late...will all users (including anons) be able to edit posts? Or will it still be restricted to admins only?
[20:18:14] <ebernhardson> we are soon trialing edits by autoconfirmed users on, enwiki and ruwiki
[20:18:26] <ebernhardson> (i think, lemme double check which ones)
[20:18:43] <ebernhardson> yes those three
[20:19:27] <ebernhardson> i believe we are shipping that wednesday after the train deploy
[20:19:28] <ebernhardson> ?
[20:20:40] <superm401> ebernhardson, don't see why not. All blockers are done.
[20:20:56] <Negative24> legoktm: I believe that is T90670
[20:21:54] <legoktm> ebernhardson: why not allow new users+anons? are there technical blockers?
[20:21:58] <thedj> what about indexability etc ?
[20:22:10] <thedj> google i mean
[20:22:28] <superm401> legoktm, no. It's just that it's not that common you need to edit someone's post, so it's questionable if anons need that right.
[20:22:37] <superm401> But it's fully configurable, so we can revisit that at any time.
[20:22:51] <superm401> Similar to why we don't let anons move pages.
[20:22:57] <quiddity> 10 minute warning
[20:22:58] <superm401> Yes, it needs to be done, but not that often.
[20:25:01] <legoktm> superm401: I don't like the idea of new software implementing social restrictions that didn't exist before.
[20:25:41] <legoktm> superm401: it's a wiki, we adapt in response to issues. anons can't move pages because of huge page move vandalism problems...we don't have that issue with anons editing other people's posts in a malicious manner
[20:26:23] <thedj> actually, speaking of search engines. what page is the 'canonical' content page of a discussion. It's Topic page ?
[20:26:37] <ebernhardson> thedj: yes, the Topic: page
[20:26:39] <superm401> A wiki does not mean that all decisions have to be reactive. However, this configuration is not set in stone.
[20:26:46] <thedj> since google punishes you for content duplication.
[20:27:26] <ebernhardson> thedj: wikipedia isn't treated like everything else in wikipedia, they have more employee's working on the search rankings of wikipedia in google than we have working on wikipedia
[20:27:40] <thedj> ebernhardson: can i change the indexability of a topic ?
[20:27:50] <ebernhardson> thedj: i dunno really
[20:28:12] <superm401> thedj, no.
[20:28:13] <thedj> probalby not possible for LQT right now, but gonna be important for a wider deploy
[20:28:40] <dannyh> what are the issues that might come up?
[20:28:42] <superm401> ebernhardson, that might be a slight exaggeration. ;)
[20:28:54] <thedj> currently entire talk pages are marked as NOINDEX by users.
[20:28:56] <ebernhardson> superm401: i mean in terms of # of engineers
[20:29:00] <superm401> Still
[20:29:19] <jayvdb> hi, I am looking for a flow dev to help mentor GSOC project - Implement Flow support in Pywikibot
[20:29:26] <thedj> but that usually happens with the header templates. And if topics are the new 'canonical' urls, then that's where the indexability needs to be set.
[20:29:34] <thedj> or possibly inherited to.
[20:30:11] <superm401> jayvdb, let's talk about it in #wikimedia-collaboration afterwards.
[20:30:17] <thedj> think discussion pages of arbcom cases for instance. all hidden from search engines by default to limit off wiki real life impact for individuals.
[20:30:41] <dannyh> thedj that's a good question, it hasn't come up for us before -- we'll have to talk about it
[20:30:48] <superm401> thedj, but yeah, if you're aware of any specific indexing problems or ranking problems, please report them.
[20:30:58] <quiddity> legoktm, There has been some discussion about this before, including this small survey (in July 2013) where a number of editors made good points about regularly needing to fix the sandbox-type-experiments by unregistered users:
[20:31:00] <superm401> Generally, it should work, since everything should be accessible with no-JS
[20:31:44] <legoktm> quiddity: thanks
[20:31:52] <superm401> At least all content is readable.
[20:32:18] <legoktm> quiddity: oh heh, I particpated in that discussion :P
[20:32:38] <quiddity> Ok, time is up for this hour.
[20:33:01] <quiddity> Much thanks to all participants, including the Collaboration Team, and the many people who brought up good questions.
[20:33:05] <superm401> Yep, thank you.
[20:33:09] <superm401> We'll be in #wikimedia-collaboration as usual.
[20:33:09] <dannyh> yes, thank you!
[20:33:11] <thedj> dannyh: i made a ticket
[20:33:13] <jayvdb> superm401: ok.
[20:33:18] <quiddity> The Logs will be posted on meta, soon.
[20:33:20] <dannyh> thedj good, thanks
[20:34:11] <thedj> ok, my concerns are mostly addressed for an LQT rollout. just wanted the team to know that.
[20:34:25] <quiddity> Thanks again, for asking the most (and good) questions. :)
[20:34:42] <dannyh> thedj oh, good! thanks
[20:35:01] <dannyh> thedj talk to you later
[20:35:26] <thedj> i would also still advise to spend a shit load of time on your communication strategies around this, but I guess that after the last two years of various mishaps, this shouldn't come up again... :)