Working Group Minutes/EWG 2013-11-18

Attendees

IRC nick Real name
iandees Ian Dees
pnorman Paul Norman
shaunmcdonald Shaun McDonald
TomH Tom Hughes
zere Matt Amos

Summary

  • Code4OSM
    • shaunmcdonald gave some feedback about the top level design of the site.
  • Redesign branch
    • TomH wanted some feedback on whether turning the side bar into a side "box" makes the front page too "map centric" and doesn't offer opportunity to explain the community / wiki aspects of the project.
    • Some discussion of what meaningful and measurable metrics would be.
    • Some concern over whether the partials render for changesets is no longer suitable for "heavy" use when examining potential cases of vandalism.
    • Some discussion about whether something like show me the way would be a good "demo" to put on the front page, but deferred decision until later.
    • In general it seems like no one had any strong feelings against the side box, but all were slightly uneasy that it potentially detracts from the "community and data" message.
  • Osm2pgsql benchmarking
    • pnorman reported: A hetzner ex-40SSD box with no major tuning gets ~7 MT/s baseline (unordered as import). Clustering on gist(way) gets 8 MT/s. Clustering on osm_id gets ~7.

IRC Log

17:35:33 <zere> minutes of the last couple of meetings: http://www.osmfoundation.org/wiki/Working_Group_Minutes/EWG_2013-10-11 and http://www.osmfoundation.org/wiki/Working_Group_Minutes/EWG_2013-11-04
17:35:46 <zere> as usual, please let me know if something needs changing.
17:36:22 <zere> doesn't look like gravitystorm is around, and i have to apologise for not getting my actions done.
17:36:47 <zere> however, i'm starting to try and get some high level documentation done on code4osm
17:37:32 <zere> and if anyone would like to help out with writing some of the text, or making the site look better, it would be greatfully appreciated.
17:48:05 <zere> is there anything that anyone wants to discuss? apologies from me, again, i'm mildly zombified from jet lag.
17:48:13 <shaunmcdonald> I've just had a read of the site, it's first time I've seen it.
17:49:49 <shaunmcdonald> I'd be tempted to have some higher level groupings of running a clone of osm, vs using a data dump.
17:50:29 <shaunmcdonald> e.g. the only people interested in using the planet dump would be those having a clone of OSM.
17:50:58 <shaunmcdonald> I'd be tempted to have a brainstorm on the main layout of the site, then fill in the details which will be easier.
17:53:24 <shaunmcdonald> Oh it's interesting the way that you are importing the read me info.
17:53:42 <zere> ok. what i am trying to work towards is a site which pulls in documentation from other projects, with as little as possible extra "glue" documentation to bring it all together.
17:54:15 <zere> i think it's much easier for people if the documentation comes with the code. but when it's presented on a website it doesn't have to be that way.
17:54:35 <zere> in fact, it's not just the README - it'll pull in all the *.md files in a project.
17:54:54 <zere> ... it's just that i've only added a couple yet, and they only have the README.
17:59:44 <zere> shaunmcdonald: do you have any thoughts on the main layout of the site? i "took inspiration" pretty directly from getbootstrap, which perhaps isn't suitable as it's documentation for a single project. i'd be very interested in alternative ideas
17:59:48 <shaunmcdonald> On https://github.com/zerebubuth/code4osm.org there's the command you need to run to do the generation missing, same with the bundler.
18:04:29 <zere> you mean in the documentation for the site itself?
18:10:26 <pnorman> morning
18:11:59 <shaunmcdonald> zere: Longer term I'm thinking it'd be good to be able to ask the user what they'd like to do or why they are here, with some default groupings or responses. When the user answers, they get presented with the list of things that's relevant for the person, thus thinning down the options. It'd be similar to the savings selectors you get for choosing a savings account.
18:12:14 <shaunmcdonald> Yeah for generating the site itself.
18:13:43 <TomH> are we meeting? or is zere midatlantic or something?
18:14:13 <shaunmcdonald> TomH: we started the meeting 45 mins ago :-)
18:14:31 <shaunmcdonald> TomH: zere is looking for feedback on https://github.com/zerebubuth/code4osm.org
18:15:36 <TomH> err so have we changed the time then?
18:17:14 <zere> yeah, we pulled it 30 minutes back a few weeks ago - in between the EU and US DST changes
18:17:21 <zere> shaunmcdonald: is this better https://github.com/zerebubuth/code4osm.org/blob/master/README.md ?
18:17:40 <TomH> that obviously passed me by then
18:18:34 <zere> pulling it a full hour back was considered too early, but leaving it at 6pm UK was also making it tough for people. basically, there's no good time for everyone and i'm always happy to move it around and see what works.
18:19:05 <zere> shaunmcdonald had suggested i send out a ping reminder 1h before the meeting, and i've been trying to do that.
18:19:21 <zere> s/ping reminder/ping reminder on this channel/
18:20:15 <TomH> am interested in getting peoples high level thoughts on the redesign PR
18:20:15 <TomH> not bothered about details and bugs etc - those can always be sorted out
18:20:15 <TomH> but I have a fundamental concern, given that the sidebar went AWOL by default, that the frontpage is now too map-centric and I'm interested in whether others share that view
18:20:19 <shaunmcdonald> zere: that read me works better.
18:23:03 <zere> TomH: you mean it now looks like OSM is just a google clone (maps, search, routing) and doesn't do a good job of showing that it's a whole community project for editable geodata?
18:23:44 <TomH> zere: well that's roughly my worry, yes
18:23:47 <shaunmcdonald> TomH: the search used to go to a lat/lon, whereas now it goes to an osm object, which could be bad for big relations, and clones of OSM.
18:24:08 <shaunmcdonald> Where's the make a donation button gone?
18:24:46 <zere> yeah, and it's tough to find space in that layout for things like SOTM banners
18:25:07 <pnorman> Are we moving from code4osm to the redesign branch?
18:25:19 <TomH> shaunmcdonald: yes I noticed that today, but that's an issue we can deal with (it also makes it hard to look at multiple results)
18:25:30 <zere> TomH: i'm looking at redesign.apis.dev.openstreetmap.org, and the "side box" saying "welcome to OSM" is still there by default... has that changed?
18:25:35 <TomH> zere: I assume under the search - jfire showed an example of a site announcement there
18:25:45 <TomH> zere: when not logged in, yes - vanishes when logged in
18:26:36 <zere> hmmm... well, i guess once you're logged in we should have had sufficient time to make the point that it's a community project for editable geodata ;-)
18:27:08 <zere> and i found the "donate" link - it's tiny in the lower right next to the attribution.
18:27:47 <zere> but, as TomH says, it seems possible to put something elsewhere on the page if we wanted to draw attention to donations for a particular purpose.
18:27:54 <zere> pnorman: yeah, i guess so ;-)
18:27:59 <zere> #topic redesign branch
18:28:15 * zere apologises for zombification again
18:28:39 <pnorman> I think it should be evaluated on a) how it will attract visitors and b) how it will do at visitor to mapper conversions
18:29:48 <zere> do you mean "how it will attract visitors" or "how attractive it is to visitors"? because the former seems like more of an SEO problem
18:30:29 <pnorman> a) how it will retain visitors who get to the site
18:31:08 <zere> and (b) is always a tricky metric - the best metric will be how it does at visitor to map edits conversions, but that's hard to measure without doing A/B testing and i don't think anyone is suggesting that.
18:31:46 <pnorman> well yes, I'm not sure we can measure these metrics, but I think we want metrics that could at least in theory be measured
18:32:34 <TomH> well my main concern currently is whether people are generally in favour of it, before I start looking at things in detail
18:32:45 <zere> and metrics which are useful to the project. i would argue that having users is not directly useful to the project unless they are contributing; traces, notes, edits, etc...
18:33:08 <TomH> though I suspect it's such a big change that my normal line by line review will be impossible
18:33:53 <pnorman> zere: well I did say visitor to mapper conversions, which implies they're doing some form of editing
18:36:40 <zere> on the stylistic issue of not having a full height side bar, i'm pretty indifferent. it looks to me like some sort of info box that can be closed, when it's not, but i'm not opposed to the general idea of having more screen space devoted to the map as long as we're able to present information about the project in a way that people notice. in that respect, having it look like an info box is perhaps an advantage.
18:38:01 <pnorman> I'm also not worried about the change from a menu to links in the top-right. The items in the sidebar menu always seemed somewhat random
18:38:39 <zere> pnorman: sure, i wasn't suggesting you thought user=mapper... i'm only trying to think aloud about the difficulty of measuring actually useful things - just like you said ;-)
18:39:51 <pnorman> Can we have jfire add a SOTM banner or a donation drive banner to the site so we can see where it'd go and how it'd look?
18:44:19 <TomH> there's a picture somewhere in the comments with a downtime notice in place
18:44:37 <zere> looking through the thread for the PR, i think i'm now slightly in favour of the "side box" over the old "side bar". the way it works on / adapts to mobile browsers seems more natural.
18:46:08 <pnorman> I'm dumping some canvec data into the redesign api so I can look at the changeset pages in the context of an import, I'm worried they don't work for "heavy" use
18:47:13 <zere> TomH: is this the one? https://f.cloud.github.com/assets/98601/1534608/c6fb207c-4c97-11e3-8885-1151aa13ec78.png
18:48:00 <TomH> yep
18:48:53 <iandees> ( pnorman: i'd love to work with you on making tools that are explicitly designed to help you find importers )
18:49:26 <pnorman> iandees: the tools to find a changeset are different, this is examining a large changeset for vandalism, imports, wtf did they do, etc
18:49:57 <pnorman> imports are just the easiest way for me to get a large changeset on an API without much data :)
18:49:59 <iandees> TomH: i like the PR in general. it's a step in a direction -- i agree that it's probably away from showing the community and data aspect of OSM front and center, but the current page doesn't really do that very well
18:50:50 <iandees> pnorman: i know, i'm suggesting improving finding changesets and viewing them.
18:50:55 <iandees> but that's offtopic right now :)
18:51:41 <zere> do you think that some of the other work on "timelines", ppawel's OWL-based history, etc... would improve the showing of community / data front and center?
18:53:50 <iandees> show-me-the-way seems to impress people whenever i show it to them
18:54:01 <zere> i guess we're missing the link to "download the data", but since a first time visitor to the site is pretty unlikely (imho) to download 30+ GB of data without first having some knowledge of how they're going to read / process / use it, i don't think it's much of a loss.
18:54:48 <iandees> showing facebook/twitter-esque timelines would be another way to show that there are other people in the OSM world, but i don't know what it would look like on the home page
18:55:09 <iandees> especially for a logged-out visitor
18:55:33 <zere> show-me-the-way is all javascript, so it would be possible to integrate that pretty easily... might be interesting to enable that by default... or would that be too confusing?
18:56:10 <pnorman> it'd be interesting to show live edits on the map for thecurrent bbox, but I think that's out of scope here
18:56:51 <iandees> it'd probably be a little confusing, but the text saying "____ edited ___" is pretty descriptive
18:57:06 <zere> sure, it's orthogonal to the "side box vs side bar" issue that TomH brought up originally.
18:57:12 <iandees> maybe it could go away when the user grabs the map and starts panning
18:57:30 <shaunmcdonald> could do a mini version of live.openstreetmap.fr on the front page to highlight all the changes that are happening?
18:57:45 <iandees> shaunmcdonald: that's what show-me-the-way does
18:58:07 <shaunmcdonald> I'm off home now, getting hungry :-)
18:58:24 <iandees> enjoy food :)
18:59:26 <pnorman> I'm overall happy with it, I just want to test-drive the browse/changeset pages under heavy usage
19:00:23 <zere> in summary on the side box vs side bar thing - it sounds like no one has any strong feelings against the side box, but we're all slightly uneasy that it potentially detracts from the "community and data" message.
19:00:31 <zere> is that accurate?
19:01:13 <iandees> yes, with the point that the detraction is basically the removal of a link to the wiki (i.e. there wasn't a whole lot to begin with)
19:01:33 <zere> because i think we've had some ideas here already to address the "community and data" message which are orthogonal to, independent of and non-blocking w.r.t the redesign PR.
19:01:40 <iandees> agreed
19:02:04 <zere> and there are some (if RichardF is lurking) who will consider the removal of the link to the wiki to be a feature ;-)
19:02:23 <iandees> :)
19:02:48 <pnorman> I don't want to try to write by committee rightnow, but could rewording the intro pop-up text to mention data and community address this?
19:02:51 <zere> although it hasn't really been removed, has it? just moved, along with links to our major supporters, to the /about page?
19:03:26 <zere> haha, ok... no: there's no link to the wiki on /about either ;-)
19:03:28 <TomH> and /help
19:03:36 <TomH> ok, so only /help
19:04:44 <zere> oh, ok. i thought that the "help" link went straight to help.osm.org... i hadn't realised there was an intermediary page
19:08:44 <zere> pnorman: agreed. i think we're possibly talking about some stuff which can go in as improvements either pre- or post-merge, in the usual way, without EWG needing to try and "committee" it.
19:09:20 <zere> i'm thinking that this is pretty much talked out... was there anything else anyone wanted to discuss?
19:10:10 <zere> pnorman: btw, nice work with the benchmark baseline. i'm looking forward to seeing the effects of the changes.
19:12:35 <pnorman> zere: yes I have numbers
19:13:03 <pnorman> a hetzner ex-40SSD box with no major tuning gets ~7 MT/s baseline (unordered as import)
19:14:00 <pnorman> clustering on gist(way) gets 8 MT/s. clustering on osm_id gets ~7.
19:14:14 <pnorman> I am betting based on previous results that ORDER BY way will *decrease* performance.
19:17:23 <zere> yup. it looked like it would be doing way more index page updates based on the map data you showed before.
19:17:40 <pnorman> this is purely for rendering, not updates
19:18:16 <zere> ok, well, same thing - it'll have to look at way more index pages with those long, thin envelopes.
19:18:46 <zere> thanks to everyone for coming! i hope to see you next week :-)