Working Group Minutes/EWG 2014-05-12

Attendees

IRC nick Real name
iandees Ian Dees
shaunmcdonald Shaun McDonald
zere Matt Amos

Summary

  • Meeting minutes
    • zere apologises for not having them ready for today.
    • ACTION zere to get the minutes ready for the next meeting.
  • AWS / "big resources"
    • "if resources weren't an issue - what would we be interested in doing?"
    • AWS are interested in helping but need motivation. one motivation would be a "big" task that will get them press or that they can write whitepapers about.
    • They have, so far, not shown much interest in helping out with smaller-scale stuff, despite peoples' efforts to engage with them.
    • Ideas:
      • mkl raised the idea of automatic imagery analysis, trying to compute if "something" is present on a tile which can then be compared to OSM to target areas where there's something in the real world, but nothing in OSM.
      • Another suggestion is multi-lingual maps (would require much more rendering power than currently available).
      • Rendering the full history over time as a series of tiled videos rather than still images, so you can pan anywhere in the world and view any period of the map evolution there.
      • Global routing analysis to spot problems with the data.
      • GPX/routing analysis along the lines of ris' "That Shouldn't Be Possible".
      • A system that stores all object's history in S3 as static files.
      • Offline Map tiles/Geocoding/Routing to the Amazon Kindle Fire or something similar.

IRC Log

17:31:01 <zere> hi everyone. i'd normally link to the last minutes for approval, but the last few meetings have clashed with travel and i've not got around to that yet. apologies.
17:31:13 <zere> #action zere write up minutes for next meeting
17:31:38 <zere> if you want to view the raw logs, they're all archived here: http://matt.dev.openstreetmap.org/osm-ewg/2014/
17:34:04 <zere> we only have two agenda items (i think perhaps there was a third, but i can't find any note i might have made for myself about it).
17:34:38 <zere> they are; 1) AWS, 2) meeting time
17:34:41 <zere> #topic AWS
17:35:42 <zere> AWS have offered to support some sort of very large computation which would be of use to OSM. it should be noted that both pnorman and iandees have been in contact with the AWS folks regarding running smaller-scale stuff, without much interest from the AWS guys it seems.
17:36:06 <zere> so the message apparently is that they're only interested if it's something really, really big.
17:36:46 <iandees> i think the message is that they're interested in helping but need to be prodded into action. one of the sticks to poke them with is a big task that will get them press or that they can write whitepapers about.
17:37:36 <zere> sure, if that's a nicer way of framing it.
17:37:43 <iandees> :)
17:37:57 <iandees> my point is that they're not only interested in huge things.
17:38:54 <zere> and my point is that even though they say nice things about helping, they've not actually been particularly forthcoming to yourself or pnorman with AWS credits or suchlike.
17:39:12 <zere> but that's sort of off-topic anyway....
17:39:39 <zere> the topic for this meeting is: "if resources weren't an issue - what would we be interested in doing?"
17:41:09 <zere> mkl raised the possibility of automatic imagery analysis, trying to compute if "something" is present on a tile which can then be compared to OSM to target areas where there's something in the real world, but nothing in OSM.
17:43:12 <zere> another item which has been suggested is multi-lingual maps (would require much more rendering power)
17:43:22 <zere> anyone have any other ideas?
17:43:33 <zere> don't hold back - the larger and crazier the better :-)
17:43:55 <shaunmcdonald> The image comparison idea is probably the best I’ve seen so far, and meets the AWS in that it can be done and turned off, and done again at a later stage if needed.
17:44:17 <shaunmcdonald> Whereas multi-lingual requires a longer comitment from Amazon.
17:45:24 <zere> how about rendering the full history over time as a series of tiled videos rather than still images, so you can pan anywhere in the world and view any period of the map evolution there? not very useful, but pretty cool.
17:45:47 <shaunmcdonald> Oh that could be interesting
17:45:58 <shaunmcdonald> Or a slider to show the map at any time
17:46:10 <zere> shaunmcdonald: yeah, that's a good point. i have a feeling in a couple of years it'll all be about client-side rendering anyway, so perhaps not such a long-term commitment :-)
17:46:33 <zere> yeah, that's what i meant - a time-bar at the bottom where you can zoom to any place and any time and start watching.
17:46:55 <shaunmcdonald> Global routing analysis to spot problems with the data.
17:47:05 <zere> once the full history is done, it would only really need to be bulk updated for new styles.
17:47:13 <zere> yup...
17:47:34 <zere> not sure that's AWS-scale, though? doesn't keepright do a bunch of that already on fairly modest hardware?
17:48:08 <shaunmcdonald> True, I suppose it’s a case of improving the keepright interface.
17:49:34 <zere> there's probably a bunch of analysis we could do (like ris' TSBP) on the gpx traces...
17:50:59 <shaunmcdonald> zere: yeah, that would be a good one. I wonder if Amazon have GPS traces that they can provide to help with that?
17:52:56 <zere> i suppose they must have a fleet of vehicles. but they don't do their own deliveries to residential addresses (except with UAV microcopters), so it might just have coverage of the major arterial networks.
17:53:01 <zere> iandees: any ideas?
17:53:50 <shaunmcdonald> Maybe if they get the tracking from the delivery companies, bit far fetched though.
17:56:27 <iandees> it's not particularly huge-scale, but it'd be nice to have a system that stored all object's history in S3 as static files
17:56:33 <iandees> (or maybe not that useful, even)
17:56:46 <iandees> and/or tiled raw OSM data
17:59:16 <zere> history, different from the /history call?
18:00:14 <iandees> instead of querying the database, all single-object calls would be a file on S3. i realize it's not the most intensive query on the DB, but it's an idea :)
18:01:02 <iandees> e.g. instead of /node/1234/history doing a SELECT * FROM nodes WHERE id=1234;, it would redirect to /node/1234.history.osm or whatever in S3
18:02:17 <shaunmcdonald> What happens when someone changes their display name? All their contributions need to be updated.
18:02:35 <zere> api 0.7 happens, and we get rid of that stuff :-)
18:03:35 <zere> i guess the same is true of redactions... not that we have many, but it's fairly important if data needs to be redacted that we're able to expire it cleanly from S3 or whatever caches.
18:05:12 <shaunmcdonald> Yup
18:06:04 <shaunmcdonald> Offline Map tiles/Geocoding/Routing to the Amazon Kindle Fire or something similar
18:07:43 <zere> hmmm.... the effort there is mostly on the client side, isn't it? doesn't really need all that much by way of "cloud" resources?
18:11:18 <shaunmcdonald> true likely more client side, beyond the processing and storage of the data.
18:11:42 <shaunmcdonald> Which would then be re-usable across other mobile devices.
18:17:14 <shaunmcdonald> How about analysis to produce estimates for building outlines etc.
18:19:27 <zere> sounds interesting, but what do you mean? analysing the aerial imagery to extract building outlines, or just to try and guess what %age of the tile is covered by buildings?
18:20:00 <shaunmcdonald> Both
18:29:54 <shaunmcdonald> I need to get going home shortly.
18:30:51 <zere> ok, just before we close - shaunmcdonald finds it more convenient to have the meeting an hour earlier. i don't mind... pnorman, iandees, would anyone find it easier/harder?
18:30:55 <zere> #topic meeting time
18:36:48 <zere> no interest? ok... we can come back to it another time.
18:36:54 <zere> #topic AoB
18:37:05 <zere> was there anything else anyone wanted to discuss?
18:41:08 <zere> ok. thanks to everyone for coming - see you next week :-)