14:33:28 #startmeeting Docs Working Group aka DaWGs 14:33:28 Meeting started Tue Sep 29 14:33:28 2020 UTC. 14:33:28 This meeting is logged and archived in a public location. 14:33:28 The chair is samccann. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:33:28 Useful Commands: #action #agreed #halp #info #idea #link #topic. 14:33:28 The meeting name has been set to 'docs_working_group_aka_dawgs' 14:33:39 #topic opening chatter 14:33:46 Ok who's around to talk docs today? 14:35:36 felixfontein gundalow abadger1999 andersson007_ ? 14:36:20 hi! 14:36:22 I'm partially around 14:36:37 #chair felixfontein 14:36:37 Current chairs: felixfontein samccann 14:36:48 you get furniture anyway :-) Could be a quiet day 14:37:07 thanks ;) 14:37:50 Hi 14:37:58 ooo more furniture! 14:38:04 #chair gundalow 14:38:04 Current chairs: felixfontein gundalow samccann 14:38:07 * gundalow sits 14:38:18 today's agenda - https://github.com/ansible/community/issues/521#issuecomment-696842671 14:38:52 To start with... 14:39:01 #topic The Case of the Disappearing Redirects 14:39:41 So yesterday, when someone pointed out a post on reddit, we discovered that the redirects went missing (the http redirects). So you couldn't go from 2.10 module to 2.9 module - you'd get a 404 14:40:10 Hi, i am on foot as usual at this time, do can’t take part in 14:40:12 #info republished the redirects last night. Not sure why they disappeared to begin with (http redirects for docs module pages) 14:40:32 ok thanks andersson008_ 14:40:41 woopsie! heh... andersson007_ 14:41:21 So I think abadger was going to create a cron job to keep an eye on things to see if this failure happens again. 14:42:12 we republished the devel docs and that didn't trigger it. And the nightly build went off per usual last night (again just devel) and that didn't trigger it. 14:42:16 did the .htaccess files vanish? 14:42:23 I believe so, yes 14:42:29 If there are URL we can test I can add some basic testing 14:43:01 to uptimerobot.com 14:43:14 This is the url that works for latest - https://docs.ansible.com/ansible/latest/collections/community/general/online_user_facts_module.html#ansible-collections-community-general-online-user-facts-module 14:44:01 i.e. https://docs.ansible.com/ansible/2.9/collections/community/general/online_user_facts_module.html should be a redirect, not a 404 14:44:20 This url - https://docs.ansible.com/ansible/2.9/collections/community/general/online_user_facts_module.html#ansible-collections-community-general-online-user-facts-module should trigger a redirect to https://docs.ansible.com/ansible/2.9/modules/online_user_facts_module.html#ansible-collections-community-general-online-user-facts-module 14:44:26 and https://docs.ansible.com/ansible/latest/modules/online_user_facts_module.html should be a redirect and not a 404 either 14:45:21 gundalow - once you have a test in place, let me know and I'll trigger a republish for 2.10. That's about the only thing we didn't try last night 14:47:06 unless someone has other ideas on how to troubleshoot/detect if this happens again? I'll move on in the agenda 14:48:53 #topic redirecting module index pages 14:49:11 how long does the republish take? (and do you have timestamped logs for it?) 14:49:23 (OK, I've added monitoring to make sure https://docs.ansible.com/ansible/latest/collections/community/general/online_user_facts_module.html exists and contains some text) 14:49:30 * gundalow -> afk 14:49:42 The republish for 2.10 takes about 1.5 hours 14:50:23 but I'll wait for abadger1999 to be online before I do it, incase it is the trigger. 14:50:34 #action samccann to republish 2.10 14:51:11 :+1: 14:51:30 on module index pages - we've gotten two requests: 14:52:11 #info requests to redirect the old module index pages to the collection index page - https://github.com/ansible/ansible/issues/71927 14:52:31 #info also some folks miss the ability to search on the all module index page to find a suitable module 14:53:28 it's that 2nd item that needs a bit of thought. Folks used to be able to search all modules page based on words/terms they are looking for, which might not be the module name but would be in the module short description. I can't think that we have anything like that now 14:55:36 I'm wondering if there is a way to recreate it that wouldn't make the docs build take forever? 14:55:42 I also did that in the past 14:56:18 I guess we could create a huge page which links all plugins from all collections, with FQCNs in them (or create one per plugin category) 14:56:34 it would need FQCN and the short description, yeah. 14:57:10 one per plugin category might be the way to go as it's closest to what we used to have. 14:57:13 like a combination of all per-collection indexes 14:57:18 Morning 14:57:20 yep 14:57:27 hey! more furniture!!! 14:57:27 morning abadger1999! 14:57:33 #chair abadger1999 14:57:33 Current chairs: abadger1999 felixfontein gundalow samccann 14:58:50 I think a republish of 2.10 probably will kill the redirects but since we hadn't done that before we didn't test it yesterday 14:59:11 I can check the rsync options in Jenkins to confirm 14:59:20 so you think that's the culprit? 15:00:56 I think it probably will break the redirects but if we didn't do it last time then we have another culprit too 15:01:50 ok we'll test after this meeting then 15:03:00 abadger1999 what do you think about the idea of recreating the 'all module index' page ^^ ? Doable without a lengthy delay in the docs builds? 15:05:20 It's not natural but i think we have all the data in memory at one time so we could do it without much additional cost 15:06:10 #info all module/plugin pages could be created w/o lengthening the docs build much 15:06:31 #samccann - open an issue to track the idea and collect feedback on whether we want to do it or not 15:06:39 Where to put the page and what to put on it at probably the biggest questions I'll need answered 15:07:29 yeah that's why I figure I'll open an issue (on antsibull?) to decide if we want to do it and what to put on it. I expect they'll be kind of ugly pages. That's a lot of text when you get the FQCN and short description going on it. 15:07:44 I'd say next to the collection index, i.e. same directory 15:08:05 For end users, the raw size of the page will get unwieldy but they're asking for it so.... 15:08:25 yeah maybe we'd call it 'searchable plugin index' or something. With a note on the top that this is to search crtl-F style or something 15:08:34 15:09:04 btw, https://docs.ansible.com/ansible/latest/collections/community/general/index.html no longer (compared to 2.9 docs) groups modules by subdirectories 15:09:51 Note, there's an index feature built into sphinx but i don't know if it would work well for this or if we'll want to make our own thing by templating something in rst ourselves 15:09:51 yes, we don't really have an option for modules by category so to speak anymore 15:10:48 I know someone mentioned missing the by category once... but I only heard that one once, vs the 404 problem and the all module index - those have been repeated a couple of times each at least. 15:10:50 the main categories are collections now :) 15:10:52 We were thinking at some point there should be some sort of metadata to tag plugins. Then we could construct lists based on those tags 15:11:07 based on galaxy tags? 15:11:41 But there's a lot of questions to answer before doing that.. where are tags stored? Who curates the tags? 15:12:04 yep 15:12:33 collection maintainers curate them, I guess 15:12:35 Are Galaxy tags per collection? If so, they won't work for community general as each plugin could have a different category 15:12:40 in c.g the tags are subdirectories right now :) 15:12:46 (same for c.n) 15:13:00 abadger1999: galaxy tags are per collection 15:13:13 15:13:15 they are specified per collection in galaxy.yml 15:13:51 I'm assuming those subdirectories aren't in the info that antsibull gets when pulling down the collection docs? 15:14:07 #action felixfontein add docs links to galaxy.yml in c.g, c.n which point to https://docs.ansible.com/ansible/latest/collections/community/*/ 15:14:23 samccann: the subdirectories are there 15:14:47 they are more painful to get from ansible-doc though, since you need to follow symlinks yourself 15:15:04 (in case that didn't change yet) 15:15:05 ah. 15:15:37 #action samccann to check the web analytics for how many visitors to the all module index page vs the module-by-category pages 15:15:50 antsibull-changelog has some logic for determining them: https://github.com/ansible-community/antsibull-changelog/blob/main/antsibull_changelog/plugins.py#L103-L121 15:15:55 I'd like to see how popular the others are before we chase down that possible solution 15:16:53 #info antsibull-changelog has some logic for determining subdirs which act as categories for c.g and c.n : https://github.com/ansible-community/antsibull-changelog/blob/main/antsibull_changelog/plugins.py#L103-L121 15:17:00 just so we don't forget :-) 15:17:45 should we move on in the agenda? 15:18:43 Yep 15:20:02 #topic bitflipping the release 15:20:15 This could be a quick topic as I forgot the last status :-) 15:22:00 anything we can cover on this one, or should I just move on? 15:23:55 ok will move on. 15:24:14 #topic ansible-test: allow version_added tests for collections 15:24:27 #info https://github.com/ansible/ansible/pull/69291 15:24:49 huh, is that still on the agenda? 15:24:51 felixfontein: this one was in the 'post 2.10' agenda and I forgot why. Is it something we should discuss? 15:24:54 HAHAH yep 15:25:16 it is essentially waiting for mattclay to implement per-collection config files for ansible-test 15:25:54 I have another PR for added_version (and removal versions) though: https://github.com/ansible/ansible/pull/71679 15:26:14 and one for fixing some missing collection names for removal versions in the docs: https://github.com/ansible/ansible/pull/71735 15:27:08 the former one makes sure that no patch releases are mentioned as version_added, and no minor/patch releases as removal versions 15:27:27 i.e. it screams at people who violate semver and document their violation :) 15:27:41 hmm... so I couldn't add a module in a patch release, it has to be a minor release? 15:27:55 yes 15:27:56 But Ansible itself can add it in a patch release (e.g. 2.10.1)? 15:28:32 ansible isn't using semver (and is excluded from this). and ansible doesn't add features in patch releases either. (it removes in "minor" versions though.) 15:29:05 ?? I'm pretty sure Ansible 2.10.1 will have new modules 15:29:17 unless I'm confused (which is no surprise :-) 15:29:24 ah I thought you meant ansible-base 15:29:30 that won't get new features in patch releases 15:29:37 ansible itself does, but only from collections 15:29:56 ah so collections and ansible-base don't add new things in patches, but Ansible the package will 15:30:05 (from the collections) 15:30:13 yes. but as opposed to collections, neither ansible nor ansible-base uses semver 15:30:25 not at all confusing. ;-) 15:30:47 collections are supposed to be using semver, which disallows new features in patch releases, and disallows removals (or breaking changes in general) in minor and patch releases 15:30:49 So on your three PRs, will you take them to the core meeting to get some movement on them? 15:31:04 yep, I already put them on th eagenda 15:31:15 yeah semver makes sense. My brain's been rewired to semver so I'm forgetting how the other two work ;-) 15:31:33 cool! 15:31:38 I hope ansible and ansible-base will eventually switch to semver 15:31:45 fingers crossed! 15:32:05 that would mean having 3.0.0 though, and it sounds like some people don't want that (yet) :) 15:32:13 Gonna do a quick open floor since we're at the 1 hr mark 15:32:18 heh 15:32:23 #topic Open Floor 15:32:41 anyone have anything they want to bring up? favorite PRs? docs issues? general docs comments? 15:34:12 ok gonna close this one out then. Thanks everyone! 15:34:37 thanks samccann for moderating the meeting! 15:34:41 #endmeeting