SamatsWiki

OSMPlaces is meant to be an open equivalent of Google Places or Facebook Places, based on OpenStreetMap data.

I don't know what a "place" is exactly, but I am going to loosely define it as something that has a name. Places have names because people want to refer to them; because they can be referred to, they should be more easily found, and have an unambiguous computable representation so we know we are referring to the same thing.

From that, two goals follow:

Ian Dees (iandees) thought up a fantastic pronunciation for OSMPlaces — awesome places.

Thanks to Laurence Penney (lorp) for many ideas! See IRC discussion below.

Contents

  1. Phases
    1. Phase 1
      1. URLs
      2. Page content ideas
    2. Phase 2 ideas
  2. IRC log
    1. 05 Jan 2011
  3. Other reading

Phases

Phase 1

URLs

[INFO] Please add more examples in this section!

URLs path components should follow the schema of:

Since country, state, and county lines generally rarely change (unlike cities), they are great for disambiguating a URI.

For example:

Some things are larger than counties. So, they can move up a level. For example:

I am unsure how to describe smaller things. Such as buildings, which may not be unique. Some ideas… they might be part of a larger entity that is unique. For example:

An open question is how to represent things not part of these larger entities, which includes many commercial places. Such places wouldn't necessarily be unique. Fast food places, for example, are not unique to cities. Perhaps by calculating their nearest named intersection?

This translates well to REST-ey type things.

If you'd like to fetch RDF for a place, http://OSMPlaces.com/USA/New-Mexico/Socorro/Oscura-Army-Airfield.rdf would fetch RDF triples (in RDF/XML) of associated OpenStreetMap tags. Perhaps .rdf.n3 for Notation3.

To fetch GeoJSON for a location, http://OSMPlaces.com/USA/New-Mexico/Socorro/Oscura-Army-Airfield.json would fetch JSON that could easily be used by a web app. Returned GeoJSON could return a polygon representing the area (if available) and a centroid point, or just the point itself.

Page content ideas

A mini-map displaying the place on OpenStreetMap. A link to the same location on Google Maps, Bing Maps, etc — we're the open ones, we don't need to discriminate against other services!

Link to the equivalent place on Google Places, Facebook Places, Foursquare, Yelp, etc. Again, we don't need to discriminate against other services.

A list of the tags associated with the place, including things like phone number, address, etc.

Include the contents of the relevant Wikipedia page (Wikimapia does this), for SEO/quick reference purposes.

A big "Know more about this place? Contribute to OpenStreetMap" link.

Phase 2 ideas

A lightweight "place" editor, that lets you easily add key/values to places. As-is, OpenStreetMap does not have a lightweight editor for quickly adding information, and the "gap of execution" of doing so is high. Say a user wants to add a phone number or the opening hours of their favorite pub: at the moment, they need to load up Potlatch or JOSM, which are much too complicated for all but the most dedicated of users. As general editors used by many of OSM's userbase, it doesn't make sense to make them any simpler. OSMPlaces "places" editor would be a simple, data-entry editor for beginning users.

IRC log

05 Jan 2011

17:15 < Laurence> is there any reason the browse UI isn't directly editable?
17:15 -!- supermag [~Fingerpri@ti0010a380-0075.bb.online.no] has joined #osm
17:15 < Laurence> the UI on the main osm site, as in http://www.openstreetmap.org/browse/node/854053101
17:17 < xjjk> it'd be really nice if it was, actually
17:17 < xjjk> don't have to load up a full heavy editor
17:17 -!- speedevil1 [~user@226.4.113.87.dyn.plus.net] has quit [Quit: Leaving.]
17:18 -!- speedevil [1000@mauve.plus.com] has joined #osm
17:18 < Laurence> yeah, i just edited something in potlatch but the load time is a fair old while
17:18 < xjjk> this is all stuff I've been meaning to bring up on -dev
17:18 < Laurence> huge overkill for the housename -> housenumber typo i was fixing
17:18 < xjjk> also, SEO of things in OSM
17:18 < Borbus> I saw somethign strange the other day.. a cyclepath that goes straight across the middle of a roundabout
17:19 < Laurence> you mean, all browseable element pages should be linked from list pages?
17:19 < Borbus> I suppose the cyclepath should technically connect to the roundabout
17:19 < xjjk> not sure what you mean by "list page"
17:19 -!- heroid [~heroid@178.175.78.253] has quit [Ping timeout: 480 seconds]
17:19 < xjjk> but Points of Interest, relations, and areas should be findable by Google
17:19 < xjjk> Wikimapia does a great job of this
17:19 < Laurence> a page, perhaps generated by taginfo, listing all shop=butcher in the uk
17:20 < Laurence> and same for all tags
17:20 < xjjk> Laurence: I was thinking more oriented around named places
17:20 < Laurence> divided up geographically, perhaps as geofabrik divides up the world
17:20 < xjjk> call it an open/wiki Google/Facebook/whomever places
17:20 < xjjk> this has a semantic web aspect too
17:20 < Laurence> so all named places would be friendly to search bots?
17:21 < xjjk> I want to be able to refer to things canonically
17:21 < xjjk> Laurence: yes
17:21 < Laurence> there's a "download xml" on those browser pages
17:21 < Laurence> the inward links are the missing things for search engines
17:21 < Laurence> though google is perfectly able to parse planet.osm?
17:22 < Borbus> Hmm.. I've been editing OSM for 6 hours today
17:22 < xjjk> Laurence: why should search engines need to parse XML...
17:22 < xjjk> instead of just making the 'browse' pages friendlier/prettier
17:22 < Laurence> because they digest it more nicely? i thought you were proposing a more semweb approach
17:22 < Laurence> agreed, the browse pages could be prettier
17:23 < xjjk> Laurence: well, OSM is already setup for semweb
17:23 < xjjk> pretty well
17:23 < xjjk> except for a few needed things...
17:23 < xjjk> canonical URLs being one of them
17:23 < Laurence> and more like, say, a Yelp page about a place
17:23 < xjjk> Laurence: bingo
17:23 < xjjk> that should be SEO and user-friendly
17:23  * speedevil resists the urge to stab people.
17:24 < xjjk> and as you mentioned, should be able to add tags (say, add a phone number or opening hours)
17:24 < xjjk> without having to launch Potlatch or JOSM
17:24 < Laurence> certainly
17:24 < xjjk> if you want to do that, now... it's like 3-4 clicks
17:24 < xjjk> 3-4 confusing clicks that'd you'd never find unless you were an experienced OSM user
17:24 < Laurence> i was thinking of making custom OSM UIs for some streets around Bristol
17:25 < xjjk> hrm. osmplaces.com is available
17:25 < Laurence> e.g. Gloucester Road and Stokes Croft are much loved by residents and shop owners
17:25 < Laurence> could even be a linear UI, not geographic
17:25 < Laurence> and the editing would directly edit OSM elements
17:26 < Laurence> (.org nicer)
17:26 -!- Blackadder [~Blackadde@87-194-16-233.bethere.co.uk] has quit [Quit: heading for the next unmapped place=]
17:26 < Laurence> each editor would have to create their own OSM account of course
17:26 < xjjk> Laurence: from a semweb standpoint, it'd be nice if OSM had canonical friendly URLs for things
17:27 < Laurence> but once the buildings and housenumbers are in, there are very few relevent changes to the geography
17:27 < Laurence> such as?
17:27 < xjjk> Laurence: e.g., say I want to refer to Disney Land
17:27 -!- supermag [~Fingerpri@ti0010a380-0075.bb.online.no] has quit [Remote host closed the connection]
17:27 < xjjk> something like osmplaces/r/347/Disney-Land
17:27 < Laurence> who are you btw, do we know each other?
17:28 < xjjk> Laurence: I'm Samat; I don't think I know you?
17:28 < xjjk> Laurence: in RDF objects are referred to by URLs
17:28 < xjjk> and the idea is that these URLs be canonical
17:28 < Laurence> ok, i don't think we know each other
17:28 < Laurence> would the final text portion need to be = the object name?
17:28 < xjjk> i.e. http://samat.org/#me would refer to me
17:28 < xjjk> Laurence: no
17:28 < xjjk> but it'd be nice for user-friendliness/SEO
17:29 < Laurence> is 347 the relation id?
17:29 < xjjk> yes
17:29 < xjjk> that doesn't need to be in there either
17:29 < xjjk> but another obstacle here is that OSM doesn't have unique identifiers for "objects" or "things"
17:29 < Laurence> would be nice to enforce the canonical as the name, if the element had a name
17:29 < xjjk> there isn't a way to keep track if something started out as a node
17:29 < xjjk> then graduated to an area
17:30 < xjjk> and to a relation
17:30 < xjjk> especially if the name changes
17:30 < Laurence> using .../r .../n and .../w isn't so bad, is it?
17:30 < Laurence> oh yeah that
17:30 < xjjk> Laurence: yeh, I just mentioned that as an easy-to-do first step
17:30 < Laurence> i brought htat up again on the list the other day
17:30 < xjjk> along with the names
17:30 < xjjk> names are user-friendly
17:30 < xjjk> IDs are not
17:30 < Laurence> i've been upgrading many nodes into building outlines
17:31 < xjjk> yeah
17:31 < Laurence> best practice seems to be to use the node as a part of the new outline
17:31 < xjjk> Laurence: yep
17:31 -!- diegows [~diegows@motherbox.xtech.com.ar] has joined #osm
17:31 < xjjk> that's not immediately useful, though
17:31 < Laurence> not very nice if someone else edits 
17:31 < doctau_> Laurence: there was a UID discussion a few months ago (tagging list?) about this :)
17:31 < xjjk> OSM doesn't have a way to expose history
17:32 < Laurence> and each parser has to implement its own method of tracking that
17:32 < xjjk> in a semantic/machine computable-way anyway
17:32 < Laurence> oh i didn't see the UID discussion
17:32 < xjjk> meh, I need to catch up reading the OSM mailing lists
17:32 < Laurence> we shifted to UIDs years ago in my other web project, and haven't looked back
17:32 < doctau_> but it got messy quickly. If I tag a building which is a shop with a UID and the shop moves, was the UID for the physical building or the shop?
17:32 -!- doctau_ is now known as doctau
17:33  * diegows is thinking about the new license terms and have question about accepting it with the public domain
17:33 < xjjk> doctau: _should_ be the shop
17:33 < Laurence> indeed, i moved Cafe Kino to its new location 40 yds away the other day
17:33 < Laurence> and felt a bit dirty doing it
17:33 < xjjk> the shop is the entity people care about
17:33 < xjjk> doctau: people would also care if that entity changed locations, for example, though that's another tangent
17:33 < xjjk> nobody cares about the building
17:33 < Laurence> uh huh... some do
17:34 < xjjk> well, most don't =/
17:34 < doctau> a lot of people do care about the building
17:34 < Laurence> architectural historians love buildings and tracking changing usage
17:34 < Laurence> for the street sites i mentioned and may yet set up, i am interested if the community manages to keep track of building occupancy in the years after a new Tesco comes in
17:34 < xjjk> yeah, I'm going to take a perspective that more people care about the shop than the building
17:34 < xjjk> also: the building isn't moving
17:35 < Laurence> that requires tracking of buildings and its social function
17:35 < xjjk> nor is its address changing/etc
17:35 < xjjk> it's more permanent
17:35 < xjjk> it doesn't really need a UUID in the first place
17:35 < xjjk> it's not ephemeral
17:35 < doctau> xjjk: that's fine, but it gets complicated if the shop moves into two new buildings instead of their old one, or similar
17:35 < xjjk> doctau: mmm yeah
17:35 < doctau> I think that's where the UID discussion got bogged down - what exactly are you tagging with the UID
17:36 < Laurence> i don't see your point related to UIDs ? some people will want to refer to that building, and in many many cases the WayID will last for years, and hence be useful as an external ID
17:37 < doctau> the building may not move physically, but it can get changed from a node to a closed way (and so keep a UID across it)
17:37 < Laurence> the point i don't get is how you (xjjk) relate permanence to a lack of requirement for UIDs
17:37 < Laurence> doctau, that's certainly a great advantage of UIDs for buildings
17:38 < doctau> I think a lot of people would like UIDs, but implementing them in a way which isn't too easy to break when people who don't understand it all edit is hard
17:38 < Laurence> do you think many people move building outlines to new locations, if the business moves?
17:38 < Laurence> i agree, it happens a lot with nodes (i did one myself)
17:39 -!- ajturner [~ajturner@209.155.228.129] has quit [Quit: ajturner]
17:42 -!- iandees is now known as Guest3430
17:42 -!- Guest3372 is now known as iandees
17:45 -!- hawke [~BGTDOM\am@71.87.27.171] has quit [Ping timeout: 480 seconds]
17:45 < crschmidt> xjjk: YCbCr JPEG-in-TIFF done right has no visual difference from an uncompressed TIFF, and has an up to 130x space saving.
17:45 < crschmidt> xjjk: If you've got the space and disk bandwidth for uncompressed imagery, great, more power to you.
17:45 < crschmidt> But if you're dealing with country-wide NAIP, worldwide landsat, large area .15M resolution imagery -- uncompressed is not practical.
17:46 < crschmidt> And the speed cost from MrSID is *huge* -- compared to even JPEG decompression you're talking about a 10x CPU usage difference.
17:47 < crschmidt> For small areas or small amounts of imagery, this is all moot
17:48 < crschmidt> uncompressed tiff is fine
17:49 < crschmidt> for large collections of imagery, you have to do *something* smart -- and YCbCr 75% JPEG with a 1-bit mask lets you create seamless JPEG-in-TIFF mosaics at a 50-100x space savings, meaning you can actually fit a lot more imagery on disk (and get at it a lot faster)
17:50 -!- DeathBaba [~Alex@178.126.183.246] has quit [Quit: Leaving.]
17:51 < xjjk> sorry, disconnected
17:51 < xjjk> crschmidt: hrm, not familar with YCbCr JPEG-in-TIFF...
17:52 < xjjk> Laurence: I can't really defend things outside my own use cases
17:53 -!- randart [~Mike@78.32.112.209] has quit [Ping timeout: 480 seconds]
17:54 < xjjk> but for buildings, roads, etc: these are all things still in OSM
17:55 < xjjk> and my pointing out that they tend to be more permanent is that regular OSM IDs as-is are fine for purposes using building information
17:55 < xjjk> there isn't information being lost
17:56 -!- randart [~Mike@78.32.112.209] has joined #osm
17:57  * xjjk hits his Internet
17:57 -!- Aison [~Foobar@zux182-175.adsl.green.ch] has quit [Quit: Verlassend]
17:59 -!- zorko [~guille@201.252.29.204] has joined #osm
18:01  * ToeBee envisions Al Gore h itting back
18:04 -!- randart_ [~Mike@78.32.112.209] has joined #osm
18:05 -!- randart [~Mike@78.32.112.209] has quit [Ping timeout: 480 seconds]
18:13 < crschmidt> xjjk: Well, it doesn't really matter, since you've already outright rejected JPEG as useless.
18:14 < xjjk> crschmidt: my notion of JPEG is useless, yes
18:14 < xjjk> if the same limitations don't apply to YCbCr JPEG/TIFF then obviously I'm interested
18:14 < xjjk> how does it get around blocking artifacts?
18:15 < crschmidt> What blocking artifacts? All things will have pixelation when you go in far enough.
18:15 -!- emacsen_ [~serge@pool-151-200-236-10.washdc.btas.verizon.net] has quit [Ping timeout: 480 seconds]
18:15 < crschmidt> Part of the design of JPEG is that you can tweak quality levels to the point where there's no visible difference.
18:15 < xjjk> there aren't blocking artifacts with "lossless" formats
18:15 < xjjk> JPEG2000 lossless, PNG, etc
18:16 < xjjk> and when you zoom in... the differences become visible
18:16 < xjjk> am I wrong for wanting to push the limit of 1m/pixel imagery?
18:19 -!- zorko [~guille@201.252.29.204] has quit [Quit: Leaving.]
18:20 -!- harry-wood [~harry@91.125.73.114] has left #osm []
18:20 -!- tjikkun [~tjikkun@2001:7b8:356:0:204:bff:fe80:8080] has quit [Ping timeout: 480 seconds]
18:23 -!- randart [~Mike@78.32.112.209] has joined #osm
18:24 < crschmidt> xjjk: Honestly, you sound like someone who wants things but has not put any effort into trying them.
18:24 < crschmidt> There *are* blocking artifacts with lossless imagery if you go beyond the source resolution.
18:25 < crschmidt> And again, JPEG is designed such that you can adjust quality to the point where you have practically no visible differences.
18:25 -!- springmeyer [~springmey@173-14-245-214-washington.hfc.comcastbusiness.net] has quit [Quit: springmeyer]
18:25 -!- randart_ [~Mike@78.32.112.209] has quit [Ping timeout: 480 seconds]
18:25 < xjjk> crschmidt: I'm glad you think you're being helpful resorting to personal attacks
18:26 < Komzpa> meooow.
18:26 < xjjk> pretty sure the first thing I asked when you mentioned it was "do you have more info?"
18:27 < Komzpa> why do you call JPEG useless? it can compress better than BMP. :3
18:28 < xjjk> Komzpa: you're directing that to crschmidt, right? those are words he put into my mouth, not my own
18:29 -!- NEOhidra [~spectrum@hyperion.hot-chilli.net] has left #osm []
18:29 < Komzpa> miau. I usually type in IRC with my fingers, not mouth. 
18:30 < Komzpa> but, anyway, good night to everyone :)
18:31 < xjjk> meh, I don't have any GeoTIFFs lying around to compare...
18:35 -!- ndim [hun@moooo.n-dimensional.de] has quit [Ping timeout: 480 seconds]
18:35  * pnorman tosses xjjk a 1.4 GB GeoTIFF
18:36 < xjjk> I have a 40 GB GeoTIFF on another machine I don't have access to ATM
18:36 < pnorman> What I suspect you'd find in this case is that you're making z19 tiles from a z18.5 source
18:36 < xjjk> pnorman: you know, I've never bothered to check/calculate that... I always let gdal2tiles.py pick the zoom
18:36 -!- tjikkun [~tjikkun@195-240-122-237.ip.telfort.nl] has joined #osm
18:37 < pnorman> It isn't perfect
18:38 < balrog-k1n> i remember some program refusing to talk to me at all if i wanted it to load, or generate, a tiff bigger than 2GB
18:38 < xjjk> balrog-k1n: thats most programs
18:39 < balrog-k1n> xjjk: how do you deal with 40GB tiffs then?
18:39 < pnorman> very slowly :)
18:39 < xjjk> use the programs that do work
18:40 < xjjk> it was a lot of work to find them
18:40 < ToeBee> hmm I assume width=* tags are supposed to be in meters?
18:40 < pnorman> xjjk: it seems a *lot* faster running gdal2tiles on this machine then the old cpu
18:40 < ToeBee> the FAA diagrams list runway width in feet
18:40  * balrog-k1n assumed that tiffs just couldn't handle bigger files (integers would overrun) so no other program would work with them anyway.. perhaps this was wrong
18:40 < pnorman> ToeBee: yes
18:40 < xjjk> balrog-k1n: there's a BigTIFF standard
18:40 < balrog-k1n> ToeBee: i think you can add a unit at the end
18:41 < xjjk> programs need to explictly support it
18:41 < xjjk> gdal does
18:41 < balrog-k1n> i see
18:41 < ToeBee> heh there are 809 "width=narrow" tags
18:42 < ToeBee> units at the end are not used at all
18:43 < ToeBee> like maybe 100 out of 484,000 tags
18:43 < xjjk> ToeBee: how many airports in KS had airport diagrams?
18:44 < xjjk> in NM I only found 1 airport
18:44 < ToeBee> I didn't look for all of them. But at least... 3? 4?
18:44 < xjjk> which was a military base, so attaching this information isn't useful for most people
18:44 < ToeBee> yeah one of them is Forbes field
18:45 < ToeBee> but mid continent in wichita, Salina municipal, the Topeka municipal one and Kansas city all had one
18:45 < ToeBee> kansas city is technically in MO... but still
18:46 < ToeBee> oh wow Manhattan regional has one
18:46 < pnorman> ugh, ogr2osm is failing with a KeyError
18:46 -!- Tensho [~kvirc@5ad1cfaf.bb.sky.com] has quit [Quit: KVIrc 4.0.2 Insomnia http://www.kvirc.net/]
18:46 < gregory> hmm, I thought great_britain.osm.bz2 was about 380MB
18:47 < ToeBee> pnorman: it's my bug!
18:47 < pnorman> ToeBee: No, I tried with and without the -1
18:47 < gregory> oh, nevermind the download has stopped at 400M
18:47 < ToeBee> oh. Then it's your fault :p
18:47 < pnorman> IIt's on line 594
18:48 < ToeBee> typo in the array key?
18:48 < pnorman> for nodeID in segmentNodes[segmentID]: with segmentID=-37720
18:48 < pnorman> No, I've had it work on other files okay, just not a couple
18:49 < pnorman> I wonder if it's related to it being negative
18:49 < pnorman> No - works on other - numbers.
18:50 < xjjk> eh. is -37720 important?
18:50 < xjjk> whatever it is
18:50 < Laurence> xjjk: can you point me to how wikimapia handles google well?
18:50 < pnorman> dunno wtf it is honestly
18:51 < xjjk> pnorman: if you want to hack around it...
18:51 < xjjk> change that line to for nodeId in segmentNodes.get(segmentID)
18:51 < iandees> Laurence, what do you mean "handles google well"?
18:51 < xjjk> Laurence: yes... give me a second
18:51 < xjjk> Laurence: er, yes, actually
18:51 < xjjk> what DO you mean handle google well?
18:52 < xjjk> I was thinking you meant give you a search query for which Wikimapia is a top result
18:52 < xjjk> I've found a couple dozen of these in the past few months
18:52 < xjjk> it's what I get for mapping the middle of nowheres of America
18:52 < pnorman> xjjk: then it errors with for nodeID in segmentNodes[segmentID]:
18:52 < xjjk> er, sory, is that a different line or the one you changed
18:52 < pnorman> same
18:53 < xjjk> pnorman: you're supposed to change the line
18:53 < xjjk> instead of doing a dictionary access, call the get method
18:53 < xjjk> the former will raise an exception, the latter will return None/null which should just let the program continue
18:53 < pnorman>   File "ogr2osm.py", line 594, in <module> for nodeID in segmentNodes.get(segmentID): TypeError: 'NoneType' object is not iterable
18:53 < xjjk> Laurence: Google "Oscura Airfield"
18:54 < xjjk> Wikimapia first hit
18:54 < xjjk> nothing in Google's results is OSM...
18:54 < xjjk> and IMHO, it should be
18:55 < xjjk> Laurence: of course, this particular example is contrived because it's a military airfield in the middle of nowhere in restricted airspace
18:55 < xjjk> but that's not the point: it shows up when people search for it
18:55 < pnorman> xjjk: What OSM site would you say it should be?
18:55 < xjjk> how is that a bad thing
18:55 -!- gregory [~gregory@188-222-50-43.zone13.bethere.co.uk] has quit [Quit: Leaving]
18:56 < xjjk> pnorman: the openstreetmap.org browse page, probably
18:56 < xjjk> display the tags
18:56 < xjjk> a mini map
18:56 < xjjk> etc
18:56 < pnorman> xjjk: What OSM site would you say it should be?
18:56 < pnorman> whoops
18:57 < pnorman> xjjk: I fixed that error by adding in a check on the line before. if segmentNodes.get(segmentID):
18:57 < xjjk> pnorman: that works too
18:58 < pnorman> fixed it in one other location, now to check the data
18:58 < Laurence> thanks xjjk, interesting
18:58 < Laurence> and damn groupon for spamming my life
18:58 < xjjk> Laurence: haha, have you unsubscribed?
18:59 < xjjk> the unsubscribe e-mail video is *GREAT*
18:59 < Laurence> an ad on wikimapia
18:59 < xjjk> oh
18:59 < xjjk> Laurence: I have an (completely unsubstantiated) feeling that Wikimapia's better SEO is how they recruit mappers
18:59 < Laurence> ok, i don't think i want to sub just to see the unsub ad!
18:59 < xjjk> one of the ways, anyway
19:00 < xjjk> if I was the self-driven kind of person to search for a middle of nowhere airfield
19:00 < Laurence> yeah, it's quite a big deal i agree
19:00 < xjjk> and found info about it
19:00 < xjjk> and could have a way to contribute information about it
19:00 < xjjk> I would
19:00 < Laurence> we should be competing on coverage of obscure places that the general public don't know are mapped
19:00 < xjjk> I don't do it in Wikimapia's case because it's Wikimapia
19:00 < xjjk> Laurence: that's another issue entirely
19:00 < xjjk> that I won't talk about
19:01 < xjjk> but, back on the topic of information that is already there
19:01 < xjjk> that OSM doesn't expose well/could expose better
19:01 < xjjk> wondering if I should register osmplaces.com and just do this myself =/
19:01 < Laurence> sorry, i meant that we are competing - but the public don't know it without good seo
19:01 -!- randart_ [~Mike@78.32.112.209] has joined #osm
19:01 < xjjk> said airport would be osmplaces.com/airports/Oscura_Army_Airfield
19:01 < xjjk> and on that page, a minimap and list of tags
19:01 < xjjk> and big link CONTRIBUTE TO OSM
19:02 < Laurence> i've thought something like osmplaces is inevitable at some point, and that it would get picked up by OSMF and run in parallel
19:02 < Laurence> with warnings about changing ids when there's info attached
19:02 < xjjk> Laurence: I don't know how to tackle the ID problem
19:02 < xjjk> other than not attempt to tackle it at all
19:02 < iandees> it doesn't need to be an ID
19:02 < iandees> in that case
19:02 < pnorman> I've wondered about how to handle that - how can you point at the particular object that represents something in the real world.
19:02 < iandees> you could search by name
19:03 < xjjk> iandees: in what case?
19:03 < xjjk> iandees: but, what if the name changes
19:03 < Laurence> are you saying use "Oscura_Army_Airfield" as the id?
19:03 -!- randart [~Mike@78.32.112.209] has quit [Ping timeout: 480 seconds]
19:04 < iandees> when you get to a wikimapia result page it's something like http://wikimapia/USA/Nevada/Area_51
19:04 < Laurence> the one xjjk just mentioned is not like that
19:04 < Laurence> http://wikimapia.org/3815512/Oscura-Aaf-Aux-Landing-Strip
19:04 < xjjk> Laurence: it could be
19:05 < Laurence> you mean some wikimapia pages are like that, or they could reprogram?
19:05 < xjjk> Laurence: er, if there were an OSMPlaces.com, that is what it could do
19:05 < Laurence> oh, i thought you were talking about programming inside wm
19:06 < xjjk> I don't care about WM
19:06 < xjjk> other than that they are doing something like tis
19:06 < xjjk> and it appears to be working
19:06 < Laurence> nor me 
19:06 < iandees> honestly you could make static pages for the entirety of OSM
19:06 < iandees> and update them weekly or daily
19:06 < xjjk> iandees: that's what I was thinking
19:06 < xjjk> just dump a bunch of static pages
19:07 < xjjk> list of tags. mini map. big link 'contribute to osm'
19:07 < iandees> accept changes into your local DB
19:07 < Laurence> the browse pages are that already - however they are ugly, not listed anywhere nice for google to find, not editable, and not commentable
19:07 < iandees> or that
19:07 < xjjk> make a huge sitemap.xml and submit to Google/friends
19:07 < xjjk> iandees: that was an idea Laurence had earlier, actually
19:07 < xjjk> editing on the browse pages
19:07 < xjjk> you shouldn't need to launch potlatch/josm to add a phone number or opening_hours to a place
19:07 < xjjk> it's a huge barrier to contribution
19:08 < iandees> yea
19:08 < xjjk> iandees: though, rather than submitting that to a local db... said thing could just be a webapp
19:08 < Laurence> could even have arrow 'nudge' keys to move things +/-5m around
19:08 < xjjk> that connects to OSM via OAuth
19:08 < xjjk> all this stuff alread exists!
19:08 < Laurence> but that's prob stage 2... :-)
19:08 < xjjk> just need to write it!
19:09 < Laurence> what did you think of the 'linear street' idea, xjjk?
19:09 < xjjk> Laurence: linear street?
19:09 < Laurence> the idea would be to present a street rather like those old routefinding books from the 19th century
19:10 < Laurence> they show a street, set in type, all straight
19:10 < xjjk> I have no idea what you're talking about
19:10 < xjjk> have an example?
19:10 < Laurence> with things either side positioned roughly in proportion to where they are if you straightened the street
19:10 < pnorman> I draw maps like that by hand sometimes
19:10 < xjjk> iandees: just wondering, do you think this is an interesting/useful idea?
19:10 < Laurence> 1 min
19:11 < xjjk> I've been trying to think of an SOTM presentation/paper
19:11 < xjjk> writing an OSMPlaces.com seems like a good idea
19:11 < Laurence> yes, it's very like how we draw by hand when we ignore minor bends and curves
19:11 < xjjk> it'd be an open competitor to Facebook Places, Google Places, etc
19:11 < xjjk> Laurence: reminds me of subway/train maps?
19:11 < Laurence> you'd need to preselect areas which are well enough stocked with POIs
19:12 < Laurence> i was thinking of launching an 'OSMsquare' with checkins and badges when i added all the shops down several local streets
19:12 < xjjk> Laurence: not a bad idea
19:12 < Laurence> yes, like subway maps
19:13 < Laurence> i came up with monitor and prefect as the two levels of 'mayoralty'
19:13 < xjjk> Laurence: I thought about doing that too
19:13 < Laurence> terms from my old school
19:13 < xjjk> except, I don't know anything about game theory
19:13 < xjjk> and I am not sure I care enough to learn
19:13 < iandees> https://docs.google.com/drawings/edit?id=12tTJfCJ_Ws55P86uubTC1qaA4AVOxnl2uUdzL2jCQvo&hl=en&authkey=CMbB0u0G
19:13 < xjjk> there is a LOT of interesting stuff going on in Foursquare
19:13 < xjjk> which is why it's successful
19:13 < Laurence> no, incentives for normal people still needs work for a non-commercial project
19:14 < Laurence> nice iandees
19:14 < xjjk> iandees: I see, interesting
19:14 < pnorman> They actually did some research with people drawing maps by hand into what gets scaled up and down by them
19:14 < xjjk> iandees: Laurence: a headache I see with this... which goes back to the UUID discussion earlier... what's a "street"
19:14 < Mike1_> somehow I don’t get at all what iandees drew there
19:14 < xjjk> a single logical way would be divided into a bunch of little different ways on OSM
19:15 < xjjk> indicating stuff like different speed limits/etc
19:15 < xjjk> I know relations are supposed to be used to group the segments together
19:15 < xjjk> but I don't do that
19:15 < xjjk> I don't think many others do
19:15 < pnorman> Renderer support for those relations is lacking iirc
19:15 < iandees> Mike1_, showing an idea for an alternative browse page
19:16 < xjjk> Mike1_: I forget the name, it's a style of map
19:16 < iandees> view the businesses along a road way
19:16 < xjjk> rather than indicate actual topology
19:16 < Laurence> http://books.google.com/books?id=l51bAAAAQAAJ&dq=Paterson%E2%80%99s%20roads&pg=PA472#v=onepage&q&f=false
19:16 < xjjk> you just list the street/way/whatever as a line
19:16 < xjjk> and points of interest along it
19:16 < xjjk> many mass transit systems/subway maps/etc are indicated like this
19:17 < xjjk> there's a name for it, I can't recal...
19:17 < xjjk> I read about it in a book about the NYC subway, haha...
19:17 < Mike1_> ah
19:17 < Laurence> this is an old UI that would work quite well for a modern street
19:17 < xjjk> Laurence: neat!
19:17 < xjjk> Laurence: very
19:17 < Laurence> i actually have that book, it's very nice
19:17 < xjjk> Laurence: this would be a nice browse/index page of sorts for such a OSMPlaces
19:18 < Laurence> it would work as a UI for any route that a routing engine found
19:18 < speedevil> Laurence: I assume you're not also known as laurencb?
19:18 < Laurence> no
19:18 < Laurence> more often Lorp online in fact
19:18 < Mike1_> what’s that strange book? oO
19:18 < Laurence> routing, 1820 style
19:19 < speedevil> k
19:19 < Mike1_> and people were supposed to be able to use that?
19:20 -!- Mike1_ is now known as Mike1
19:20 < Laurence> what idea do you propose for a UI made only paper printed using lead type?
19:20 < ToeBee> so the JOSM validator really needs to be changed to not complain about north/south/east/west relation memberships
19:21 < Mike1> Laurence: just telling people to get a map and see how to get from A to B ;)
19:21 < Laurence> it's really rather a lovely UI when you study it a bit
19:21 < Laurence> maps are very expensive to print, in 1820
19:21 < Laurence> also they get damaged easily
19:21 < Laurence> printing using lead type, much better!
19:22 < xjjk> Laurence: my RESTy gears are flowing
19:22 < Mike1> Laurence: I’ve nevery been able to find a place with verbal descriptions
19:23 < Laurence> love maps or hate words?
19:23 < ToeBee> hmm no "josm" component in trac?
19:23 < iandees> josm.openstreetmap.de
19:23 < pnorman> josm has it's own
19:23 < ToeBee> bah!
19:23 < xjjk> something like osmplaces.com/USA/New-Mexico/Oscura-Army-Airfield.html
19:23 < xjjk> want that in GeoJSON? osmplaces.com/USA/New-Mexico/Oscura-Army-Airfield.html gets you a centroid point/polygon/whatever
19:24 < xjjk> want the tags in RDF? osmplaces.com/USA/New-Mexico/Oscura-Army-Airfield.rdf
19:24 < Laurence> xjjk - what things would you def include and def exclude?
19:24 < Laurence> roads? landuse?
19:24 < xjjk> Laurence: no idea
19:24 < xjjk> Laurence: named things
19:24 < xjjk> not roads, probably
19:24 < Laurence> good answer
19:24 < xjjk> roads don't really have a "location"
19:24 < xjjk> they go from point a to point b
19:24 < Mike1> Laurence: hate people ;) People’s unique perception and memory make their descriptions quite useless for everyone besides themselves.
19:25 < Laurence> blimey Mike1
19:25 < iandees> a road's page could list the stuff on that road
19:26 < Laurence> yes, ian, i think turned into a straight vertical strip you could do quite an interesting ui on that
19:26 < xjjk> iandees: yah... in the manner Laurence described/you linked to
19:26 < Laurence> like the 1820s pages
19:27 < iandees> that'd be hard to keep SEO-able
19:27 < Laurence> how so?
19:27 < Laurence> all html5, hardly any graphics
19:27 < xjjk> iandees: how so? just a page going down
19:27 < xjjk> yes, HTML5/CSS3 rock
19:27 < iandees> true
19:27 < xjjk> can do all these rotations/etc
19:27 < iandees> well let's go guys, get it done!
19:27 < Laurence> oh shit, ok
19:28 < xjjk> iandees: would you think this is an interesting SOTM paper/poster...
19:28 < iandees> heck yea
19:28 < Laurence> my angle on the roads thing, is that they'd be great as a kind of community hub
19:28 < xjjk> go me
19:28 < iandees> but don't waste time writing the proposal, just implement it :)
19:28 < xjjk> iandees: I've been looking for an excuse to load up the OSM DB...
19:28 < Laurence> a bit like a better presented and more localized http://www.myyate.co.uk/ (which is done my a bristol OSMer)
19:29 < xjjk> hot
19:29 < xjjk> osmplac.es is available
19:29 < Laurence> heh
19:29 < xjjk> I can even have a silly web 2.0-ish URL!
19:29 -!- randart [~Mike@78.32.112.209] has joined #osm
19:29 < xjjk> the key to success
19:29 < iandees> is place.ly  available?
19:29 < Mike1> just wait for the IPcalypse :p
19:30 < xjjk> iandees: I doubt it... and I really have qualms about Libya anyway
19:30 < xjjk> it's really low that so many web 2.0 companies have gone in that direction
19:30 < iandees> yea but it's so 2.0
19:30 < xjjk> it is =/
19:30 < xjjk> place.ly is taken by someone in New Jersey
19:31 -!- randart_ [~Mike@78.32.112.209] has quit [Ping timeout: 480 seconds]
19:31 < Laurence> xjjk, shall we buy a domain each and hassle each other to get stuff done?
19:31 < Laurence> i was thinking of the ord
19:31 < Laurence> org
19:32 < iandees> you have to promise to donate at least some of the Google ad revenue back to OSM :)
19:32 < Laurence> i think that's fair using the OSM tm
19:34 < iandees> i think the neatest piece for this idea would be a customizable mini-POI editor for a way/node ID
19:34 < Laurence> xjjk, do you mind if i grab it, since you first said get osmplaces.com
19:34 < xjjk> iandees: yeah... I'd like that to be a phase 2 ideas
19:34 < Laurence> yes, that would be a cool first use
19:34 < xjjk> Laurence: you're getting .es or .com?
19:34 < Laurence> .org
19:34 < iandees> i.e. <script src="http://osmplaces.org/editor.js?node=429385"> would create a iframe or something
19:34 < xjjk> Laurence: go for it
19:35 < xjjk> I'm partial to .es
19:35 < xjjk> it's short
19:35 < xjjk> will probably get .com since it's .com and available...
19:35 < Laurence> got it
19:35 < xjjk> iandees: I am thinking a very realizable phase 1 idea is just static html pages
19:35 < xjjk> and get google to crawl it
19:35 < iandees> oh and it has to be pronounced "awesome places"
19:36 < xjjk> iandees: awesome!!!
19:36 < Laurence> heh, very cool!
19:36 < ToeBee> hah!
19:36 < Laurence> how do we get the goog to crawl? lists of all the shop=supermarket in uk, etc?
19:36 < iandees> i'm thinking about the editor because when you start out the pages will be very empty
19:36 -!- supermag [~Fingerpri@ti0010a380-0075.bb.online.no] has joined #osm
19:36 < Laurence> or do that sitemap thing
19:37 < iandees> just make pages and link between them
19:37 < iandees> then get people to link to them
19:37 < xjjk> iandees: yah.... trying to think of a simple, accessible user-friendly editor...
19:37 < Laurence> why will they be empty? do you mean lacking comments and extra stuff?
19:37 < xjjk> Laurence: sitemap thing is a start
19:37 < iandees> lacking content from the OSM data
19:37 < xjjk> iandees: er, I don't follow
19:37 < ToeBee> hmm I know a guy who does SEO for a day job. He might be able to provide some pointers once you get a basic site up
19:38 < Laurence> it would be a given, i think, that we'd import all of osm before launch
19:38 < iandees> i mean a McDonalds is going to be tagged name=McDonald's, amenity=fast_food, cuisine=burger
19:38 < Laurence> test on small areas
19:38 < xjjk> ToeBee: mehh SEO is a snake oil business
19:38 < iandees> so you get a page with a big <h1> for name
19:38 < xjjk> make simple, useful pages.
19:38 < xjjk> the world will come to you
19:38 < ToeBee> I suppose
19:38 < iandees> and then say ... McDonald's is a fast food business.
19:39 < Laurence> yes, and if  you make people see how cool it is when an area has really well tagged shops, it'll encourage tagging and use of the UI
19:39 < xjjk> iandees: mmm I'd not work that hard
19:39 < xjjk> iandees: probably, just do a search on wikipedia
19:39 < xjjk> and include a snippet on the page
19:39 < Laurence> it would be good if we could get a one-line short description instead of the whole wiki text
19:39 < iandees> that's what Mechanical Turk is for :)
19:39 < Laurence> for what shop=butcher means, etc
19:40 < iandees> oh, the OSM tags can be hand-done
19:40 < xjjk> there's simple english wikipedia
19:40 < xjjk> is there a short wikipedia somewhere
19:40 < xjjk> twitterpedia, hahah
19:40 < Laurence> i mean to explain what this tag means, by using a condensed version of what's on the OSM wiki
19:40 < xjjk> everything about the world in 140 characters or less
19:40 < xjjk> this is another site idea
19:41 < xjjk> that I will let some MBA follow up on
19:41 < Laurence> quite a cool one
19:41 < Mike1> iandees wants to make something like wikimapia?
19:41 < Laurence> you've seen http://twitter.com/booksin140
19:41 < pnorman> xjjk: It was you who I got the rename script for TMS numbering to google TMS number, right?
19:42 < xjjk> pnorman: yes
19:42 < pnorman> Should post that somewhere
19:42 < xjjk> pnorman: I need to publish that...
19:42 < xjjk> I write all these stupid scripts all the time =/
19:42 < Laurence> ok i gotta go
19:42 < Laurence> bed
19:42 < pnorman> could stick it on your wiki user page
19:43 < Laurence> good to talk to you guys, i'm known as Lorp at lot online, on OSM wiki, etc
19:43 < xjjk> Laurence: I'm SamatJain
19:43 < Laurence> at lot -> a lot
19:43 < Laurence> which part of the world?
19:43 < xjjk> pnorman: I've a gitorious/github acct I've been meaning to populate
19:43 < xjjk> USA
19:44 < xjjk> pnorman: I am not good at the software engineering side of things...
19:44 < xjjk> for example, the gdal2tiles.py thing
19:44 < xjjk> I really, really need to add a lot of error checking to that =/
19:45 < xjjk> it's really sort of lucky that I turned a single-threaded program everyone complains about is slow into a n-faster one in an evening
19:45 < xjjk> there are lot of bugs in it...
19:45 < xjjk> 2 or 3 race conditions...

Other reading

CategoryIdea

SamatsWiki: OSMPlaces (last edited 2011-09-29 22:32:50 by SamatJain)