Working Group Minutes/SWG 2011-03-04

From OpenStreetMap Foundation
IRC Name Present Apologies
_chrisfl Chris Fleming y
cmarqu Colin Marquardt
Firefishy Grant Slater y
mkl Mikel Maron y
RichardF Richard Fairhurst y
rweait Richard Weait y
samlarsen1 Sam Larsen y
stevenfeldman Steven Feldman y
toffehoff Henk Hoff y
TomH Tom Hughes y
twain47 Brian Quinion y
wonderchook Kate Chapman y
Also attending
IRC handle Name
zere Matt Amos
Blackadder Andy Robinson
filbertkm Katie Filbert
apmon_ Kai Krueger


  • Proposed by: Steven Feldman
  • Seconded by: Chris Fleming
  • Old business
  • Per arrangement last week, this meeting is chaired by Henk Hoff.
  • budget referred to board
  • Steven to revise based on board feedback, post to strategic@ for quick assent and return to board for approval.
  • tiles policy revisions were sent to board for approval.
  • no feedback from board yet. Henk to follow-up with board.
  • wikimapia was referred to board.
  • Discussion of Routing service
  • Discussion: Is routing needed and why? What restrictions should exist re: routing as a service for mappers to increase data quality, vs. consumer routing? Noted importance of not providing consumer services as they are not within the core goals of OSM, especially improving data quality.
  • routing discussion to continue on strategic@ and next week.
  • Next meeting 2011 March 11 at 1600H UTC
  • Meeting closed

IRC log

(11:01:40 AM) rweait: ** Logging begins **
(11:01:45 AM) ***RichardF 's cat (official OSM lolcat of awsomeness, as you all know) has also just turned up
(11:02:06 AM) rweait: "Stowford!"
(11:02:19 AM) toffehoff: Hello all!
(11:02:22 AM) stevenfeldman: baffled of london here
(11:02:25 AM) rweait: previous minutes for review.
(11:02:51 AM) rweait: current minutes in progress.
(11:02:54 AM) rweait:
(11:03:33 AM) rweait: I see we have a couple of guests with us today.  Any potential COnflicts of interest to declare, anybody?
(11:03:40 AM) aude|away: hi
(11:03:47 AM) aude|away is now known as aude
(11:04:03 AM) aude is now known as filbertkm
(11:04:56 AM) toffehoff: So everyone present has made him/herself known?
(11:05:13 AM) rweait: toffehoff care to start us with the agenda?  Perhaps we can have a proposal and a second for the previous minutes?  
(11:05:25 AM) toffehoff: yes. thanks.
(11:05:31 AM) Blackadder: I'm observing today
(11:05:34 AM) toffehoff: Anyone to propose or second the minutes.
(11:05:41 AM) toffehoff: Blackadder: welcome.
(11:05:52 AM) rweait: Blackadder is recognized as Andy Robinson.  Welcome Andy. 
(11:06:01 AM) stevenfeldman: propose
(11:06:02 AM) Blackadder: :-)
(11:06:12 AM) toffehoff: Thanks Steven. Anyone to second?
(11:07:41 AM) rweait: I wrote them so perhaps somebody else to confirm? 
(11:07:53 AM) ***Firefishy wasn't around last week.
(11:08:26 AM) ***filbertkm also not here last week 
(11:08:33 AM) Firefishy: _chrisfl?
(11:09:00 AM) toffehoff: I'm not sure if I can ....
(11:09:07 AM) toffehoff: But if so ....
(11:09:21 AM) toffehoff: I'm happy to second.
(11:09:31 AM) rweait: Thank you. 
(11:09:32 AM) toffehoff: Any objections?
(11:09:44 AM) rweait: Now revisions / clarifications?  Objections?
(11:09:46 AM) toffehoff: If not, minutes accepted.
(11:10:16 AM) toffehoff: With thanks to rweait for putting them up.
(11:10:18 AM) rweait: Old business?  
(11:10:40 AM) toffehoff: What do we need to discuss today.
(11:10:42 AM) rweait: results from tasks last week?
(11:10:42 AM) toffehoff: - routing?
(11:10:51 AM) toffehoff: right...
(11:11:12 AM) toffehoff: - Budgeting.
(11:11:12 AM) stevenfeldman: there are a couple of bits o/s from budget stuff, feedback from board etc
(11:11:49 AM) toffehoff: Due to plannings issues I wasn't present at the last board meeting.
(11:12:02 AM) stevenfeldman: Shall I quickly explain?
(11:12:03 AM) _chrisfl: sorry,  just distracted there - happy to second minutes.
(11:12:08 AM) toffehoff: I cannot give feedback straight from that meeting out of first hand.
(11:12:11 AM) rweait: stevenfeldman: so budget was referred to board, but no feedback yet? 
(11:12:24 AM) toffehoff: stevenfeldman: go ahead.
(11:12:24 AM) stevenfeldman: Yes feed back rweait
(11:13:18 AM) stevenfeldman: Feedback was ltd but OK - 1 suggestion was that the reporting/monitoring might vary by project, which makes sense as bigger projects will need more scrutiny
(11:13:59 AM) toffehoff: ltd?
(11:14:04 AM) rweait: stevenfeldman: do we need to add a budget item to the agenda today?
(11:14:06 AM) stevenfeldman: We will revise to allow feedback/monitoring forms to be posted on OSM wiki and referenced in the approval i.e use form A or Form B
(11:14:12 AM) RichardF: toffehoff: "limited"
(11:15:02 AM) stevenfeldman: 2nd question was re the open-ness of the process and whether it was top down or bottom up, that came from steve c
(11:15:51 AM) stevenfeldman: Discussion that I had was about our route where all submissions are via the WG's and have a personal sponsor would act as a filter on a bottom up process
(11:16:10 AM) stevenfeldman: sort of middle up
(11:16:37 AM) stevenfeldman: that was the totality of the feedback that I have received
(11:17:35 AM) rweait: now, what about other old business?  was Tile layer revision referred to board?  was wikimapia referred to board?
(11:17:57 AM) toffehoff: Anything else to add to the budgeting issue stevenfeldman?
(11:18:05 AM) toffehoff: Can we close this?
(11:18:27 AM) toffehoff: Next steps?
(11:18:43 AM) stevenfeldman: Nope, If everyone is happy to leave minor amendments to me I will make them. Then what needs to happen?
(11:20:01 AM) toffehoff: I can imagine to make the proposal final and send it to board for acceptance and further actions?
(11:20:32 AM) stevenfeldman: OK I will revise as requested by board and resubmit to them, will do next week
(11:20:36 AM) rweait: Can we put a quick discussion of the response into "Any other business" today, then have you return that to the board for approval?
(11:20:53 AM) _chrisfl: Sounds good, send out an e-mail when you've made the changes so we can have a look
(11:21:26 AM) toffehoff: ok... next...
(11:21:43 AM) toffehoff: wikimapia and tiles were also forwarded to board.
(11:22:05 AM) toffehoff: Have to get back to you all on that. Wasn't there last meeting....
(11:22:19 AM) toffehoff: Let's keep that pending on me.
(11:22:47 AM) toffehoff: Other items for today's meeting....
(11:23:07 AM) toffehoff: I have a list of issues we could talk about:
(11:23:10 AM) toffehoff: - Routing
(11:23:31 AM) toffehoff: - Setting up bounties (eg for site redesign)
(11:23:57 AM) toffehoff: - How to attract new contributors (PR?)
(11:24:11 AM) toffehoff: - Professional support for WG's?
(11:24:25 AM) stevenfeldman: I thought routing was the number 1 priority
(11:24:35 AM) toffehoff: Think it is.
(11:24:40 AM) toffehoff: Oh wait...
(11:24:56 AM) toffehoff: the other stuff was related for budget-proposals :-)
(11:25:08 AM) toffehoff: let's talk about routing then
(11:25:31 AM) apmon_: Ok, so what are the steps to take in this process?
(11:25:41 AM) ***Firefishy is ready to fill in about costs/hardware etc if asked.
(11:26:09 AM) apmon_: First, has SWG formally decided it wants to add routing to the main site and thus will support efforts to achieve this goal?
(11:26:20 AM) _chrisfl: I think we have.
(11:26:48 AM) toffehoff: I think we have concluded that routing was a welcomed feature on our main website.
(11:26:49 AM) apmon_: Secondly, have we agreed on what exactly are the aims we want to achieve with a routing on the website?
(11:27:02 AM) stevenfeldman: Can I be a voice of "are we really sure?"
(11:27:21 AM) rweait: Only if a suitable routing solution is available.  There is no need to put in a bad solution.
(11:27:29 AM) toffehoff: stevenfeldman: pls do
(11:27:48 AM) toffehoff: rweait: then what is "suitable"?
(11:27:54 AM) apmon_: So the question is then a) what is a suitable routing solution and b) how to we get one
(11:28:03 AM) toffehoff:
(11:28:13 AM) rweait: We've discussed this before.  It relates to performance, stability, dev-community ... etc.
(11:28:16 AM) _chrisfl: stevenfeldman - worth a final discussion; but we should log a decision to stop us from having the same discussion every week....
(11:28:17 AM) stevenfeldman: I don't think the case has been well made for the benefits of doing this when there are already several sites that offer routing on OSM data
(11:28:20 AM) toffehoff:
(11:29:00 AM) apmon_: toffehoff:  As I mentioned on the discussion page of the wiki, I am not sure that is quite along the right line for our current status
(11:29:01 AM) rweait: Though "why add routing" is still a valid question. ...
(11:29:07 AM) stevenfeldman: If we thought it was complicated agreeing criteria for tiles this will be even more complicated and it has a potential impact on the tech guys
(11:29:39 AM) rweait: I think the appaling debacle this week with iphove developers is a great example if why we want to be very carefull about adding services that might be considered consumer-facing. 
(11:30:34 AM) toffehoff: rweait: +1
(11:30:34 AM) rweait: We already have every "genius" with a new iPhone app scraping our resources and hurting performance for mappers.
(11:30:36 AM) apmon_: We don't necessarily have to add an external API
(11:30:54 AM) Firefishy: wiki + OSM tiles went down due to issues cause by a iPhone app over using our resources.
(11:30:57 AM) stevenfeldman: As the range of innovative uses of OSM grows we cannot incorporate them all on the front page of the web site.
(11:31:05 AM) Firefishy: I spent 1 1/2 day getting things under control.
(11:31:18 AM) apmon_: Not having resources for mappers isn't necessarily the right answere though either
(11:31:22 AM) stevenfeldman: rweait what happened with iphone devs this week
(11:31:39 AM) rweait: stevenfeldman: see firefishy above. ;-)
(11:32:12 AM) rweait: apmon_: what?  Prociding resources for mappers it the ONLY goal of SWG and OSMF. 
(11:32:16 AM) Firefishy: stevenfeldman: Their app become very popular and allowed mass downloading of map tiles. I had to hack our setup to serve them blank tiles and redirect some of their traffic off the main server.
(11:32:21 AM) toffehoff: Can we decide on: when we want routing, it should only be for internal use (read: no API)
(11:32:23 AM) Firefishy: I also blocked some of their requests.
(11:32:28 AM) apmon_: If performance really is an issue, one could even go as far as only providing the option if one is logged on
(11:32:47 AM) stevenfeldman: What do we see as the "strategic" functions of the web site?
(11:33:58 AM) toffehoff: Before we get into performance-discussions: first how do we see our services?
(11:34:09 AM) stevenfeldman: I always thought web site was for mappers and to show off the data, we can build a gallery of stellar apps using OSM that people can link out to
(11:34:25 AM) apmon_: stevenfeldman: Imho, a) to attract mappers and b) to provide feedback if the entered data is correct and sensible
(11:34:37 AM) _chrisfl: For me, I think that if people can regularly use the osm website  as a alternative for google maps and then a % of these will become editors.  
(11:35:11 AM) _chrisfl: Secondly (and related) routing can help find problems with the data.
(11:35:19 AM) rweait: _chrisfl: no.  OSM is not an alternative to google maps.  It is an alternative to NavTeq, Teleatlas. 
(11:35:47 AM) rweait: Providing Google-map-like services will kill this project in under three months. 
(11:36:22 AM) stevenfeldman: I would hesitate at trying to position OSM web site as a consumer mapping site, we don't geocode as well as the commercials (yet) and routing will be patchy where we don't have the detail
(11:36:59 AM) zere: it's worth noting that cloudmade and mapquest both offer routing APIs for developers, and should probably be considered the place to send anyone who's writing an app...
(11:37:41 AM) stevenfeldman: MapQuest are building a good consumer site with OSM data shouldn't we be pushing people to them if they want a Google alternative?
(11:37:52 AM) toffehoff: zere: +1
(11:37:53 AM) rweait: Can we recommend, and provide resources for an "experimental" routing service for mappers? It would be for logged in mappers only and stricktly rate limited (or otherwise locked up)
(11:37:57 AM) stevenfeldman: zere +1
(11:38:41 AM) _chrisfl: in that case we don't need routing; on the site. But I think that if we want too keep growing then we must widen the outlook of the web site, or rely on external sites like MapQuest to do that for us.
(11:39:07 AM) apmon_: rweait: I think that would be a reasonable compromise for now
(11:39:11 AM) rweait: _chrisfl: no, and yes, and no ;-) 
(11:39:24 AM) Firefishy: _chrisfl: to do build routing testing a would be handy.
(11:39:32 AM) Firefishy: s/build/bulk/
(11:39:53 AM) Firefishy: s/a/a service/ blah!
(11:39:59 AM) stevenfeldman: With respect (which inevitably is a bit of a not true) the web site is not too clear about objectives unless you dig a lot. Can we defer routing and schedule a serious strategic review of the web site, it's purpose etc
(11:40:08 AM) _chrisfl: Firefishy - fair point
(11:40:09 AM) apmon_: _chrisfl: by that argument we wouln't need tiles either, and I think we all agree OSM wouldn't be at the point it is now if there weren't any tiles on
(11:40:10 AM) rweait: _chrisfl: We probably do want some form of routing tool for mappers.  They will come up with cool ways to use routing that improve the data.  We don't want to crush that possibility.  
(11:40:31 AM) zere: and it's worth noting that cloudmade and mapquest both host their own editors (mapzen, branded PL/2), so it's entirely possible to gain new mappers without them even seeing the OSM website. (although that's not necessarily a good thing).
(11:41:03 AM) rweait: _chrisfl: if we want to keep growing (yes, but in the right way) we can start by retaining more of our users, or many other ways.  
(11:41:13 AM) stevenfeldman: There are some tools that we might be able to access that will test network connectivity without the load of a routing app
(11:41:38 AM) zere: for example, keepright and the cyclestreets connectivity checker.
(11:41:39 AM) rweait: _chrisfl: and we don't want to rely on the approval of any external resources for our growth. 
(11:42:00 AM) _chrisfl: rweait: agree
(11:42:37 AM) rweait: _chrisfl: ;-)  didn't mean to "pick on" you.  It was a great collection of topics. 
(11:42:58 AM) rweait: Firefishy, costs and resources? 
(11:43:05 AM) apmon_: Why do people see routing any different that tiles or a geocoder, both of which OSM has provided for a long time? It is a form to visualise data
(11:43:20 AM) stevenfeldman: If we implement a public facing routing solution and it gives poor results either because of a lack of detail in the data or a poor engine we will damage the brand and the objective of attracting more creators
(11:43:31 AM) _chrisfl: :) It's worth having the discussion as from my memory we seem to be taking a different view than previous discussions. 
(11:43:43 AM) apmon_: stevenfeldman: How is that different from showing blank tiles?
(11:43:59 AM) zere: apmon_: in my mind, because people wrote good renderers and geocoders and worked with us to get that stuff integrated. so far, noone has done that with routing.
(11:44:27 AM) stevenfeldman: apmon_ in most places we will only go blank as you drill down and we don't have more detail, peop
(11:44:31 AM) apmon_: zere: That is a different question than if we want routing or not
(11:44:55 AM) stevenfeldman: woops, people would understand that better than a poor or wrong route
(11:44:56 AM) rweait: apmon_:  and you suggest? ...
(11:45:03 AM) Firefishy: If people could take a look @ this
(11:45:32 AM) zere: not really. for it to be available there has to be the software, the skill and the will. without the first two it doesn't matter about the last one.
(11:45:41 AM) apmon_: Or
(11:46:06 AM) rweait: Firefishy: that's awesome.  Right through the channel tunnel to paris.
(11:46:20 AM) apmon_: zere: Imho all three exist (more or less) what is missing is a will by OSMF to work with those people to actually get it onto the page
(11:46:41 AM) stevenfeldman: firefishy that is very neat and performs well in uk
(11:46:53 AM) zere: Firefishy: iirc, that one takes some inordinate amount of memory to preprocess. apmon_: yep - there's the API, where's the rails code?
(11:47:02 AM) Firefishy: OK and use the 2 most likely candidate routing engines.
(11:47:03 AM) RichardF: have MQ approached OSMF asking for their routing to be on
(11:47:12 AM) zere: no
(11:47:20 AM) RichardF: ok, so that's not "the will" then is it
(11:47:38 AM) stevenfeldman: why can't we link out to other peoples' routing services rather than trying to weave into osm site?
(11:47:43 AM) apmon_: zere: the rails code is mostly a small bit of boiler plate to call the right js that already exists
(11:47:50 AM) rweait: Firefishy, costs and resources?
(11:47:56 AM) Firefishy: zere: Yes, would require a machine costing around £8000 to currently work globally.
(11:48:08 AM) zere: iirc, it's been policy so far to try not to depend on external services for the home page.
(11:48:15 AM) zere: Firefishy: 64Gb RAM?
(11:48:26 AM) Firefishy: 128GB.
(11:48:32 AM) rweait: Wow.
(11:48:41 AM) toffehoff: Firefishy: and with how many concurrent users?
(11:48:47 AM) zere: i'm sure it'll last at least 6 months ;-)
(11:48:59 AM) apmon_: zere: One of the problems is that the currently OSM is very discouraging to new developers. And thus noone is motivated to actually implement any rails code
(11:49:05 AM) rweait: And that is for performance of what, all mappers 100 routes / day or something?
(11:49:22 AM) zere: apmon_: bullshit. 
(11:49:22 AM) stevenfeldman: firefishy what level of request do you think a single server could support? Danger is that we get an overload which makes tiles API problem this week look minor
(11:49:26 AM) rweait:  apmon_: that another topic for another day.
(11:49:30 AM) Firefishy: toffehoff: routingdemo can do 100s I suspect. nroets demo can likely handle 10s.
(11:49:55 AM) RichardF: apmon_: we have exacting standards. that's why we're successful.
(11:50:11 AM) ***_chrisfl thought everyone was against routing....
(11:50:15 AM) apmon_: zere: No its not there are a number of patches, ideas and other things that would be useful, but just never get applied. Routing is imho another example
(11:50:26 AM) rweait:  toffehoff:  do we have enough to move to an action on routing or do we need to go over this yet again next week?
(11:50:26 AM) Blackadder: RichardF: Thats a nice quote
(11:50:57 AM) Firefishy: engines: = Open Source Routing Machine, = gosmore
(11:51:09 AM) toffehoff: Can we come to some sort of resolution today.
(11:51:11 AM) toffehoff: ?
(11:51:25 AM) stevenfeldman: I am a -1 on routing
(11:51:30 AM) zere: apmon_: yes, there are a number of half-finished, or even 90% finished patches. ideas don't count without someone to actually write code. what's stopping you or anyone else from finishing those things you've started?
(11:51:32 AM) rweait: What fundraising do we need to do for the 8K?
(11:51:40 AM) Firefishy: stevenfeldman: Open Source Routing Machine can do 1000s per minute
(11:51:50 AM) RichardF: toffehoff: as far as I can tell, there doesn't yet appear to be a settled consensus among this group in favour of routing on the homepage. I don't know that we can say any more
(11:51:57 AM) toffehoff: Let's first decide on what we want before talking practicalities.
(11:52:05 AM) Firefishy: Rebuilding the routing database is slow, max of say twice a day.
(11:52:07 AM) apmon_: RichardF: You can go about this in two ways: "Hey, this is a great idea. I think it is really great and would like to see it, but there are still these specific issues that need to be solved" or "Nah, it is not perfect yet, and I am not really interested in it anyway". OSM is currently more of the latter.
(11:52:12 AM) _chrisfl: I think that if we agree that we are going to implement routing (perhaps initially only for logged in users) 
(11:52:20 AM) rweait: Limited, experimental routing for mappers only +1 / -1 ?
(11:52:36 AM) RichardF: apmon_: my experience as P2 co-author suggests exactly the opposite
(11:52:37 AM) _chrisfl: then we should look at doing a "bakeoff" between the engines. 
(11:52:56 AM) toffehoff: OK, let's see where we stand on the question of rweait: experimental routing yes or no?
(11:53:05 AM) _chrisfl: +1
(11:53:17 AM) zere: apmon_: my experience with cgimap suggests exactly the opposite, too.
(11:53:36 AM) apmon_: How long did it take to implement cgimap? Nearly a year
(11:53:49 AM) rweait: abstain
(11:53:50 AM) toffehoff: We're coming to the hour.
(11:53:56 AM) stevenfeldman: -1
(11:54:00 AM) RichardF: also abstain
(11:54:14 AM) toffehoff: Looks like this needs more discussion.
(11:54:23 AM) apmon_: zere: at that is despite the author of cgimap, being a member of the TWG!
(11:54:34 AM) toffehoff: Can we summon the main questions we need to get answered?
(11:54:38 AM) zere: apmon_: check your facts. i implemented the 0.5 version in a few weeks. it sat idle for almost a year until i updated it to 0.6 and within a few weeks it was live.
(11:54:53 AM) rweait: zere, apmon_: another time please.
(11:55:20 AM) toffehoff: guys, grab a pint...
(11:55:26 AM) rweait: SWg, anything else before we wrap up?  same / same for next meeting?  11 Mar at 1600 UTC?
(11:55:45 AM) toffehoff: Next week, same time.
(11:55:53 AM) toffehoff: Main discussionl: routing (again).
(11:55:56 AM) rweait: starting with routing again?
(11:55:59 AM) zere: -1
(11:56:02 AM) rweait: Oh joy. 
(11:56:16 AM) rweait: Can we move forward some debate on the list during the week?
(11:56:23 AM) Firefishy: I'll see if I can scrape together some hardware for a backoff between OSRM vs gosmore :)
(11:56:48 AM) Firefishy: There are additional criteria, like developer support etc.
(11:56:53 AM) toffehoff: But before we go around in circles: we need some sort of discussion paper we can talk about.
(11:56:57 AM) rweait: Firefishy +1
(11:57:05 AM) stevenfeldman: I don't think I am going to be able to make the call next week may be en route home from Cambridge, apols
(11:57:14 AM) rweait: We have the chart for evaluating routers.
(11:57:25 AM) rweait: stevenfeldman: noted.
(11:57:43 AM) toffehoff: Wel...
(11:57:48 AM) rweait: Any last thoughts before we move to adjourn?
(11:57:55 AM) stevenfeldman: toffehoff before we evaluate routing should we agree on strategic purposes of web site?
(11:57:58 AM) RichardF: I won't be here next week either, will be at Sustrans conference
(11:58:14 AM) rweait: noted.
(11:58:19 AM) RichardF: thank you :)
(11:58:25 AM) rweait: Have fun!
(11:58:28 AM) toffehoff: stevenfeldman: good to take this to a more abstract discussion.
(11:58:49 AM) _chrisfl: I agree that we need to look at this from a higher level.
(11:58:59 AM) apmon_: zere: cgimap was ported to 0.6 in Mai 2009 ( It was deployed in January 2010. And only because of the Haiti earthquake and because the API was falling apart
(11:59:16 AM) toffehoff: Let's do that next week then.
(11:59:31 AM) toffehoff: What is the objective of our main website.
(11:59:44 AM) rweait: Move to adjourn
(12:00:00 PM) toffehoff: Let's adjourn.
(12:00:06 PM) toffehoff: Have a nice weekend all.
(12:00:08 PM) rweait: second? 
(12:00:12 PM) Blackadder: As an observer I'd say the discussion was getting bogged down in the detail. Strategic is best left for the really high level stuff
(12:00:15 PM) stevenfeldman: have good weekends all round
(12:00:27 PM) rweait: Thank you Blackadder! 
(12:00:29 PM) toffehoff: blackadder: thanks for the observation.
(12:00:33 PM) stevenfeldman: blackadder +1
(12:00:34 PM) rweait: Good to have you here.
(12:00:49 PM) Blackadder: pleasure, I'll try and make a few more
(12:00:53 PM) rweait: *** Logging ended ***