Working Group Minutes/SWG 2010-11-11

From OpenStreetMap Foundation

Minutes (DRAFT)


  • mkl: Mikel Maron
  • toffehoff: Henk Hoff
  • samlarsen1: Sam Larsen
  • chrisfl: Chris Fleming
  • x___: [unkown]

Discussion outline

  • Discussion about previous minutes.
    • Sam is going to prepare a document with the tile-policy recommendation. It's going to be put on the OSM-wiki first. If it needs to be fixed, we can move it to the OSMF wiki.
  • Server resources / integration
    • “Core” functionality: maintaining db, running API, generating a tile-layer
    • Prioritize services based on:
      • Essential to the OSM project
      • Supporting the OSM community
      • Promoting the OSM project
    • OSMF offering temporary hosting for innovative OSM-related projects (like FSF) in order for the project to find a more sustainable hosting solution
    • This topic needs further discussion
  • OSMF as financial umbrella organization
    • not discussed
  • Articles of association
    • not discussed
  • AOB
    • People attending meeting should identify themselves at start of the meeting

IRC log

(timestamps are in CET)

[15:12] mkl: Ok, so who's on for the meeting? 12 minutes after now
[15:12] samlarsen1: me
[15:12] toffehoff: I'm here
[15:12] mkl: cmarqu twain47 x___ ?
[15:13] mkl: anyone willing to take minutes
[15:14] toffehoff: I'll keep a log running.
[15:15] mkl: and summarize?
[15:16] mkl: ok, well let's get started then
[15:16] mkl: first issue is to move forward on the tile layer policy
[15:16] mkl: richard published an initial draft
[15:16] mkl: any thoughts on this?
[15:17] mkl: what needs clarification? what more needs to be covered?
[15:17] toffehoff: AFK
[15:18] mkl: <isthisthingon?>
[15:18] samlarsen1: unique - maybe we need to clarify this acceptance criteria
[15:18] mkl: i gave some feedback here ...
[15:19] samlarsen1: how do you overcome unique vs. competing providers
[15:19] mkl: well unique in the sense that its not covered well by an existing layer?
[15:19] mkl: it would be by applying the other guidelines
[15:20] mkl: so say more or less equal, but one had an open style sheet ... than prefers it
[15:20] samlarsen1: i.e. same type of tile service, maybe different styles - which one to choose
[15:20] mkl: samlarsen1: example?
[15:20] samlarsen1: a cloudmade style vs a (other?) style
[15:21] samlarsen1: sorry - bad example
[15:21] mkl: my opinion on the cloudmade styles is that the represent a certain kind of configurability
[15:21] mkl: so to choose from StyleEditor styles, I'd say to just go for a best example
[15:21] mkl: Midnight Commander or whatever
[15:22] samlarsen1: let's say, there is a specific cloudmade style created for environmetnal stuff, then another rendering provider also has an environmental style which comes along second, can we ditch the first in favour of the second?
[15:23] samlarsen1: even though they are not really unique
[15:23] mkl: i guess, if there was interest to switch, and it met more of the guidelines, then yes
[15:23] mkl: you are concerned about arbitrating such possibilities?
[15:23] samlarsen1: ok, so unique can be overcome in favour of the other guidelines?
[15:24] mkl: i think unique means in relation to the other styles included on osm
[15:24] samlarsen1: ok, fine
[15:25] mkl: someone is going to have to make this decision of course
[15:25] mkl: we don't necessarily need to present a solution to that to the Board
[15:25] mkl: likewise on presentation ... we can present the options
[15:26] mkl: is there anything else that we have some question on?
[15:26] samlarsen1: do you think some in OSM may pick up on the non-commercial point - e.g. adding cloudmade services?
[15:26] mkl: i'd like to know if we can move ahead to polish off the recommendation
[15:26] mkl: its only one guideline, right?
[15:26] mkl: we should make sure preferred is well defined
[15:26] samlarsen1: ok
[15:27] mkl: but also make clear that it's ok if it a tile layer doesn't meet them all
[15:27] chrisfl betrad het kanaal.
[15:27] mkl: anything else to consider?
[15:27] mkl: and
[15:27] samlarsen1: yes, should spell that out so everyone knows
[15:27] mkl: is anyone willing to take on the task of preparing this?
[15:28] mkl: preparing the policy recommendation
[15:28] mkl: i foresee we'd share that with the board, and then on osmf-talk, etc
[15:29] mkl: ok, i hear crickets
[15:29] samlarsen1: i would, do we have a doc style/format to follow or use as example?
[15:29] mkl: nope
[15:29] mkl: this is the first one
[15:29] samlarsen1: ok - i'll take it on
[15:29] mkl: it doesn't need to be anything in particular
[15:29] mkl: probably in the wiki would be fine
[15:29] samlarsen1: should it evolve from Richard's draft?
[15:30] mkl: yea, i think it could
[15:30] samlarsen1: deadlines?
[15:30] mkl: would it be reasonable for you before the next board meeting?
[15:30] mkl: or perhaps we should have one more meeting first, to review?
[15:30] mkl: either would be next week
[15:31] samlarsen1: yes, i can put it together, get a bit of feedback from yrselves
[15:31] mkl: awesome
[15:31] toffehoff: I'm back
[15:31] toffehoff: Sorry
[15:31] mkl: thanks sam
[15:31] mkl: ok, so let's move on
[15:32] samlarsen1: do you want it on OSMF wiki or OSM - i don't have access to OSMF wiki
[15:32] mkl: we have 3 fresh issues to choose from
[15:32] mkl: oh ..
[15:32] samlarsen1: i can request it i guess
[15:32] mkl: any thoughts on that toffehoff?
[15:33] chrisfl: d the first half of the meeting. Another work meeting overran, I'll have a read later.
[15:33] mkl: better to restrict access or no
[15:33] toffehoff: Putting it on OSM or OSMF wiki?
[15:33] mkl: yea
[15:33] samlarsen1: worst-case, i can put text together & pass it on to OSMF
[15:33] toffehoff: Let's put it on OSM wiki first.
[15:33] mkl: chrisfl ... just been working to wrap up tile layers
[15:33] samlarsen1: ok
[15:33] mkl: cool
[15:33] chrisfl: figured 
[15:33] toffehoff: There we still could have a bit of discussion with the community
[15:34] toffehoff: If we have to fix it in stone, we can move it to OSMF
[15:34] mkl: i imagine there will be ... i think we should present to the Board, then float to the community
[15:34] mkl: yup
[15:34] toffehoff: agree
[15:34] samlarsen1: agree
[15:34] mkl: ok, so what do we take next
[15:35] mkl: server resources and service integration (mapsomatic)
[15:35] mkl: project financial umbrella (transwiki)
[15:35] mkl: or articles
[15:35] mkl: i'd guess by the list traffic, the frist 
[15:35] mkl: also mentioned during the Board meeting yesterday
[15:35] toffehoff: let's go.
[15:35] mkl: ok
[15:36] mkl: so 2 parts to this. let's take on the first. what does OSMF provide server resources for, why, how, etc?
[15:36] mkl: what do we do currently?
[15:36] mkl: what's reasonable guidance for what we could do?
[15:37] mkl: and if it's more than our current capacity, well, how do we raise funds for servers and insure admin support?
[15:37] mkl: thoughts?
[15:37] toffehoff: First: do we want to go for 24/7 uptime garantee?
[15:38] toffehoff: This may help on what we can / cannot do on our servers
[15:38] mkl: well, i would guess that anything we're currently hosting, we have some presumption of guaranteed service
[15:38] chrisfl: So we have "core" functionality. Maintaining database, running API,  generating a tile layer.
[15:39] mkl: api is expected to be up most of the time, and if it needs to come down, tech wg announces it
[15:39] mkl: do we host search?
[15:39] toffehoff: search of what?
[15:40] mkl: what else would we conceivably host? routing?
[15:40] mkl: nominatim
[15:40] toffehoff: Ah.
[15:40] chrisfl:  Nominatim
[15:40] x____: gives  a list of what is currently hosted and how much resources it takes
[15:40] mkl: x____ : who are you?
[15:41] chrisfl: we also have the "Export" tab which is load dependant
[15:41] mkl: yea, that's been an issue in the past
[15:41] mkl: can be occassions where export doesn't work
[15:41] mkl: i've stopped relying on it
[15:42] mkl: but if we are offering it, shouldn't there be a reasonable support, somewhat not dependant on other load? dedicated service?
[15:42] mkl: i suppose mapsomatic comes in this kind of functionality
[15:42] chrisfl: yes at busy times you get a message explaining why.
[15:42] samlarsen1: can we just reduce the max extent on that service
[15:43] mkl: would we consider it one of the best, and worthy of support, under some guidelines
[15:43] chrisfl: I think it does.
[15:44] mkl: you know, whatever guidelines we drew up, there may always be a shortfall in current capacity
[15:44] mkl: we may need prioritize these
[15:44] toffehoff: like in:
[15:44] mkl: like now, site >api > tiles > nominatim > export
[15:45] chrisfl: agree
[15:45] toffehoff: -essential for OSM as a project
[15:45] samlarsen1: agree
[15:45] toffehoff: - promotional material
[15:45] toffehoff: - supporting the community (dev etc)
[15:46] mkl: then if we wanted to increase support to other services of less priority, we'd need to get estimates of what it would take in up front and recurrent costs to support, to a particular level
[15:46] toffehoff: ok
[15:46] mkl: toffehoff: yes, makes sense
[15:46] mkl: though we should dig in a bit maybe
[15:47] toffehoff: I think we don't want to become something of an ISP
[15:47] samlarsen1: i have to shoot off, i'll sort out the tile layers doc - good luck
[15:47] mkl: thanks samlarsen1
[15:47] chrisfl: I don't know when, but at some point we will run out of space/goodwill/bandwidth at UCL and may need to start adding in some external (expensive) hosting.
[15:47] toffehoff: Thanks sam
[15:47] mkl: i believe we are already using other sites ... is it imperial college?
[15:48] x____: tiles are the biggest draw on bandwidth. Most other services are not as bandwidth intense
[15:48] mkl: the other option is to ask for more gratis hosting
[15:48] x____: part of the tile traffic gets proxied via imperial, but not all
[15:48] mkl: ah ok
[15:48] mkl: i think we could find more options
[15:48] mkl: it would be up to the TWG to monitor and give indication of our needs
[15:48] x____: I think some of the planet.osm downloads get mirrored via henet, but I am not sure on that
[15:49] mkl: though to support other services, like mapsomatic, requires more servers, rack space
[15:49] mkl: and someone to admin the machine
[15:50] chrisfl: and also to support future growth of these services....
[15:50] mkl: right. the TWG already makes projections of current services
[15:51] mkl: any other services we'd evaluate?
[15:51] x____: space in the racks is imho not an absolute immediate issue at UCL, but it isn't perhaps the most "typical" datacenter...
[15:51] toffehoff: Do we want to "outsource" some of our hosting to friendly ISP's?
[15:51] mkl: ie measure against the guidelines, prioritize, estimate resources, and locate resources if it makes sense
[15:51] toffehoff: Or do we want to have everything on our own servers
[15:51] chrisfl: I think the maposmatic team volunteered to admin there own box, but there has to be co-operation with admins.
[15:52] mkl: ISPs, or friendly internet companies?
[15:52] samlarsen1 verliet het kanaal. (Quit: ajax IRC Client)
[15:52] mkl: what if mapquest offered to host osmf servers?
[15:52] toffehoff: yes
[15:52] mkl: too much conflict of interest?
[15:52] x____: in addition to the osmf servers, there are also servers run by "osmf-de" and the french chapter and some others too...
[15:52] mkl: wikipedia i believe has hosting at various corporate data centres
[15:52] toffehoff: ... and other countrie
[15:52] chrisfl: toffehoff don't know if it matters, as long as it is dedicated hardware...
[15:53] mkl: we'd need to have very clear agreements in place
[15:53] mkl: ie not gentleman's agreements
[15:53] toffehoff: agree.
[15:53] toffehoff: something in writing
[15:53] x____: what happened to the offers such as e.g. netcologne?
[15:53] mkl: i don't remember ...
[15:53] toffehoff: nothing
[15:54] mkl: so in this case, apparently mapquest offered space to mapsomatic directly
[15:54] toffehoff: this partly triggered the question of external hosting.
[15:54] mkl: woudl the alternative be to ask mapquest for possibly osmf hosting, then allocate part of that to mapsomatic?
[15:55] mkl: i would guess that mapsomatic's reluctance to take mapquest's offer
[15:55] toffehoff: not jumping for joy at that thought ....
[15:55] mkl: isn't because it's a commercial company, but because they would worry about loss of control
[15:55] mkl: me neither, just putting it out there for discussion
[15:55] chrisfl: from the e-mail train, I think they were offered a VM and couldn't get the performance they required?
[15:56] x____: was that the MQ hosting or some other hosting?
[15:56] chrisfl: we should ask them.
[15:56] toffehoff: They're now with FSF
[15:56] chrisfl:  
[15:57] toffehoff: But why would we as OSMF offer them hosting?
[15:57] chrisfl: If we wind back, it comes down to do we want to be able to offer support to OpenStreetmap projects in the same way that FSF support open source projects?
[15:57] mkl: the FSF hosting arrangement is temporary, I gather
[15:57] toffehoff: That's the right question
[15:58] toffehoff: Yes it's temp
[15:58] mkl: same way, meaning would OSMF offer temporary hosting?
[15:58] mkl: for some class of lower priority projects?
[15:59] toffehoff: On the grounds of it being a great example of what can be done with OSM
[15:59] mkl: but not essential
[15:59] toffehoff: yes
[15:59] mkl: but OSMF recognizes it could potentially be of very wide benefit
[15:59] mkl: so has interest in seeing it succeed, and in being something that could be integrated with the site
[16:00] x____: there is the dev server for "external" osm related projects
[16:00] chrisfl: exactly.
[16:00] toffehoff: .... and that they should be able to find a more sustainable solution for their hosting.
[16:00] mkl: does dev get used in this way?
[16:00] mkl: would something hosted on dev machines be integrated into
[16:00] chrisfl: but are they beyond what might be hosted on dev?
[16:00] mkl: could we guarantee reliability if a dev project? perhaps it's another step after this
[16:01] mkl: "beta"?
[16:01] toffehoff: alpha 
[16:01] x____: OWL is currently run on dev I think, but if it gets integrated into the main site, I think it is supposed to get its own "propper" machine
[16:02] mkl: oh, we're at one hour
[16:02] mkl: we could continue a bit
[16:02] mkl: or look to review where we stand now, and close, but continue out of meeting
[16:02] toffehoff: It's quite busy at my side.....
[16:02] x____: Thats something the TWG need to decide, but my impression is that anything on main page needs proper support, i.e. not the devserver.
[16:03] mkl: so perhaps some other level of server
[16:03] mkl: that's above dev, but not meant to be permanent
[16:03] chrisfl: agree - main page linking is another level. We would want hosting to be stable to integrate.
[16:03] mkl: i wonder how we'd look at routing
[16:03] mkl: perhaps for next meeting
[16:03] chrisfl: It's a good question.
[16:04] x____: routing currently runs on the dev server
[16:04] mkl: can anyone volunteer to compile minutes?
[16:04] toffehoff: routing being a good way to do some quality checking....
[16:04] toffehoff: I'll try to work something out this evening.
[16:04] mkl: thanks henk
[16:04] toffehoff: minutes I mean
[16:04] toffehoff: OK
[16:04] x____: e.g
[16:05] toffehoff: So next meeting:
[16:05] toffehoff: - Routing.
[16:05] mkl: so i don't think we've exhausted this discussion, by a long shot
[16:05] mkl: yea, I think it's related to the same question
[16:05] mkl: resource allocation
[16:05] chrisfl: no lots to think about.
[16:05] mkl: ok, so we can continue with this next time
[16:05] toffehoff: Basicly: next week continuation of this discussion.
[16:05] mkl: as well as try to wrap up tile layers
[16:05] mkl: yup
[16:06] chrisfl: sounds good.
[16:06] mkl: let's see if same time next week works
[16:06] toffehoff: x___: for the record: who are you?
[16:06] mkl: didn't have some of the key participants from last week
[16:06] x____ verliet het kanaal.
[16:06] mkl: ha
[16:07] mkl: well i'm not sure we want to have anonymous working group members
[16:07] toffehoff: ... was I to offensive .. 
[16:07] chrisfl: no.
[16:07] chrisfl: it was a good question.
[16:07] mkl: we can have an open door, but shouldn't we have to identify ourselves?
[16:08] mkl: anyway
[16:08] chrisfl: Someone reasonably knowledgeable. We maybe need to police that better...
[16:08] toffehoff: For starters.... we need to know who is participating...
[16:08] mkl: yea, i think that's fair for a Foundation meeting
[16:08] toffehoff: guys, I'm off.
[16:09] mkl: ok, me too
[16:09] mkl: thanks all
[16:09] toffehoff: closing the logging now.