Working Group Minutes/EWG 2013-07-08

From OpenStreetMap Foundation


IRC nick Real name
apmon Kai Krueger
pnorman Paul Norman
TomH Tom Hughes
zere Matt Amos


  • /changes deprecation.
    • pnorman reports no response to mailing list posting.
    • The path is open for a PR removing it.
  • cgimap pgsnapshot backend
    • pnorman reports he has a working map? call.


17:01:04 <zere> minutes of the last meeting:
17:01:18 <zere> please let me know if anything there needs changing.
17:02:05 <zere> first agenda item today is the post-hackday questionnaire:
17:03:15 <zere> this was started a few meetings ago with the aim to send it to people involved in the post SOTM-US hackday to see if we could identify any ways in which we could help new contributors become long-term contributors.
17:04:35 <zere> does anyone have anything they think should be included in this questionnaire?
17:12:00 * pnorman returns
17:12:21 <pnorman> apmon: any progress on the SOTM-US hackday questionnaire?
17:13:03 <apmon> oops, I totally forgot about that :-(
17:13:04 <apmon> Sorry
17:13:18 <zere> no worries - shall we work on it now?
17:14:36 <apmon> yes, we can do that.
17:14:40 <apmon> Where was the URL again?
17:14:45 <pnorman>
17:18:32 <apmon> Are my changes visible, or do I need to sign up to hackpad?
17:19:05 <pnorman> I see no changes
17:19:37 <apmon> I guess I'll have to register to yet another random service...
17:19:58 <pnorman> we could move it to google docs
17:20:26 * pnorman is still in waking up mode
17:21:16 <zere> yup. i'm not particularly attached to hackpad
17:21:35 <zere> or we could just dump some ideas here and organise them later?
17:22:32 <apmon> Well I have signed up now
17:23:43 <apmon> Are my changes visible now?
17:23:46 <pnorman> yes
17:24:26 * pnorman doesn't do ideas this early in the morning
17:25:34 <apmon> I am currently just writing out ideas. Not well formulated questions yet
17:26:47 <apmon> Any other topics on the ewg agenda for today?
17:27:40 <pnorman> followup on changes API.
17:27:52 <zere> yup, that's the only other item, iirc.
17:27:54 <pnorman> The followup is no one has said anything in reply to the dev@ posting.
17:28:19 <zere> in which case i think we're good to go if someone wants to submit the PR taking all that stuff out.
17:30:21 <apmon> zere: Do you have any ideas for the code sprint follow-up?
17:30:56 <zere> putting a couple in now
17:31:21 <apmon> just seeing it
17:49:27 <pnorman> zere: i should probably talk cgimap with you sometime, i'm ahead of schedule significantly
17:57:31 <zere> pnorman: ahead of schedule is an excellent place to be :-)
17:57:48 <zere> one must always assume that there'll be some bear-trap waiting at a later stage as well...
17:58:09 <pnorman> zere: as in i have a working first pass of a map? call and can point JOSM at my local dev box and everything except for the website works seamlessly
17:58:46 <zere> i should also mention that i won't be here for the next two meetings (summer vacation) - does someone want to take over running EWG meetings until the 29th?
17:59:03 <pnorman> not me, but I'll gladly take over your summer vacation :)
17:59:27 <zere> you'll have to fight me for the vacation - i feel a desperate need for it ;-)
18:00:07 <zere> great news about the pgsnapshot backend, though. congrats :-)
18:00:21 <zere> what does JOSM do if it can't load /capabilities?
18:00:29 <pnorman> dunno, since it can load it :)
18:00:38 <zere> you added that as well?
18:00:44 <pnorman> the lighttpd config I have redirects that to the main API
18:01:08 <zere> oh, ok.
18:01:10 <pnorman> when I say seamlessly I include uploads and history calls
18:01:29 <zere> i guess at some point adding /capabilities to cgimap should be easy, whenever we actually need it.
18:02:25 <pnorman> the two issues I have upcoming are a) we really could use some attribute on <osm/> to indicate what instance or server a .osm file is from
18:03:35 <pnorman> b) the nodes_from_bbox, ways_from nodes, nodes_from_way_nodes at is not optimal, and I want to refactor some of that logic down into the backend-specific stuff
18:03:46 <zere> oh, i remember i had a question about - i seem to remember using the 404 handler for a reason... something about some request header going missing otherwise. TomH: do you remember this too, or am i getting confused?
18:05:32 <zere> pnorman: not optimal, well... it depends on the backend, really. i don't want to have to re-implement every single call for every single backend. i'd rather have a small set of primitives which worked across backends and could be combined to form higher-level calls.
18:06:11 <pnorman> s/not optimal/not optimal on all backends
18:07:11 <pnorman> the above statement still needs verification with benchmarks.
18:07:41 <TomH> zere: well that looks like nginx? which we don't use...
18:07:50 <pnorman> TomH: lighttpd
18:08:12 <TomH> well same answer
18:09:12 <pnorman> I could check what happens if you ask for a missing object, that's about all I can think of that would respond differently to server.error-handler-404 and url.rewrite-once
18:12:33 <apmon> Didn't the api still use lighttpd when cgimap was introduced?
18:12:54 <TomH> maybe, but I can't remember anything about it
18:13:12 <TomH> frankly as soon as we were shot of it I wiped it from my memory
18:17:33 <zere> well, i guess if it works for you... i can't remember what the specific issue was, and perhaps it has been fixed since anyway,
18:17:41 <pnorman> double-checking on a missing node
18:18:09 <zere> my recollection was that it was some request header that went missing, rather than lighttpd tampering with the response
18:18:35 <pnorman> checking
18:18:39 <zere> like Accept or Accept-Encoding or something like that
18:19:39 <pnorman> I've found lighttpd's config documentation to be a tad lacking
18:23:12 * pnorman ponders if it could be caching
18:25:01 <pnorman> zere: I'll check incoming/outgoing headers, 404 responses from cgimap and caching
18:26:40 <pnorman> more of a general question: what will it take to get the /[node|way|relation]/# etc calls handled by cgimap on
18:38:22 * pnorman decides to merge his node with the bed node before doing some work