[Strategic] Fwd: Fw: Routing on OSM

Sun Jan 30 01:49:09 GMT 2011

I forgot to hit reply all, so forwarding my mail to the strategic list this
time.

---------- Forwarded message ----------

On Sun, Jan 30, 2011 at 12:58 AM, Milo van der Linden <milo at dogodigi.net>wrote:

> It is my personal opinion that routing should not be something the
> foundation will facilitate. Why? Because it will ask a lot of the hardware
> and it is not part of the OpenStreetMap foundation mission statement.
>

I would like to try and make a counter argument of why I think a routing
demo should be part of the osm main page, and as such an instance run on
OSMF hardware.

For one, I think it does belong to the mission statement of OSMF and
secondly I don't think it will ask unduly of hardware resources. Currently
it doesn't look like it would need any more resources than tileserving or
search, rather much less and so should be within the budget of OSMF. Also I
would be reasonably confident that if OSMF is strapped for cache, it would
get the necessary money for a routing server quite quickly if it did a
specific donation drive for a routing server.

But the reasoning for why it belongs to OSMFs mission just as much as a
tileserver and a search engine is more importantly.

I think there are at least two good reasons why it is worth for OSMF to
invest some money into routing (and these apply equally to tile serving and
search)

The core product that OSM tries to create is the raw data and that of course
is what the main aim has to be about. Ensure that the data grows as much as
possible while assuring the highest possible quality.

But if the OSM webpage only offered a link to a 14Gb compressed XML dump it
would be hard to interest many people beyond tech companies and geeks, both
together by far not large enough a group to create a good map and keep it
maintained to high standards. So show casing a few examples of what can be
done with OSM data (like a pretty slippy map, a search box or routing)
straight on the main page is quite vital to geting the broader public
interested. They have to see why it is relevant to them, why they should
care, and what they get out of OSM for spending loads of their time in
creating the data.

The second reason is quality assurance. Routing is quickly becoming one of
the most important use cases of OSM data. With companies like MapQuest,
CloudMade, Skobbler, Navmii, mapfactor and hobby projects like garmin maps,
gosmore and gpsmid, there appear to be by now  thousands if not tens or even
hundreds of thousands of people actually using OSM for routing, many of them
haven't even heard of OSM before. It offers a huge potential for recruiting
new mappers. Yet the OSM data is stil relatively poor quality for routing.

While people are for example mapping trees and lampposts or go and e.g.
trace random remote villages in countries they have never been to as  some
mappers no longer know what to add otherwise, turn-restrictions speed limits
and other routing relevant tags are distinctly lacking.

Although there are bound to be many reasons for why there is such a lack of
routing tags in the data. One I think is particularly strong is that they
aren't shown on the main mapnik map. There is a reason why it was such a big
step to get rendering updates down from the weekly update cycle to the
minutely diffs and people immediately start complaining if the rendering
falls behind. People want instant gratification of seeing that the work they
have done "shows up on the map". There currently is no equivalent for
routing tags and so there is little incentive to add them.

I am fairly confident that if OSM would include a router on the main page,
people would start adding turn restrictions, access restrictions or  speed
limits, if that is what it takes to ensure the routes they try out no longer
show up incorrectly. Routing on the main page could be a mayor Q/A tool and
that surely is within the core scope of OSMF.

I think one example of what highlights the quality problem for routing
particularly is the connectivity. Take a look at e.g.
http://tools.geofabrik.de/osmi/?view=routing&lon=-0.15948&lat=51.53711&zoom=8
which
shows the number of roads that are nearly connected but not quite. Most of
them look perfectly fine on the rendered map, but for a router they cause
completely wrong routes. So even in well mapped areas, there many such
connectivity bugs.

Hopefully many more of these problems would be fixed, if more people would
on top of haveing a look at the rendered map after uploading their changes
to check everything is correct also try out some routes through the area
they added and verify that it works for routing.

In general, there are so many great QA tools already, but they are all
distributed accross various sites so that hardly anyone knows about them or
bothers visiting these extra sites, so imho it is of strategic interest to
integrate some of these more tightly into the main OSM stack, to ensure the
data is of the highest possible quality.

Routing is becoming such an important use of OSM data, imho it deserves some
resources (and don't think it would be hard to get them) to help improve the
data for it. It would also address one of the biggest wishes mappers have to
improving the OSM webpage.

I also think it should be of core interest to SWG of which activities and
services OSMF should support and fund to grow and improve the data the
fastest and most efficiently.

Kai

>
> There are a lot of routing initiatives with openstreetmap, some are
> commercial, some are opensource, some are considering to one day become
> open. Besides that, pgrouting, the postgresql/postgis routing engine is
> picking up speed and as opposed to gosmore, pgrouting is ready for
> turn-by-turn navigation, but it also can handle travelingsalesman, A* and
> other advanced routing techniques. I often asked myself, why are there so
> many routing initiatives and why aren't they joining forces to create one or
> two kick-ass routing systems? I cannot figure it out.
>
> A role the OSMF might take is see if there are sponsors particularly
> interested in providing funds for routing, then facilitate in bringing
> developers together and let them work something out with only one real
> demand: That the routing-engine is completely opensource. But as I said
> before it doesn't fit the OpenStreetMap foundation mission statement in my
> opinion.
>
> Last time I dug into the gosmore fileformat, I noticed that it doesn't
> maintain reference to openstreetmap way id's, this makes it impossible to
> look up streetnames that could generate output like "go straight on 5th av
> for 3 miles, turn left on central square, go straight for 1 mile on 12th
> street, turn right on main". gosmore simply generated one linestring. But
> this was 6 months ago and things might have changed.
>
> Anyway, my answer would be no, and I would prefer if we can look into
> getting strategic up to steam on more foundation related matters first.
>
> Kind regards,
>
> Milo van der Linden
>
>
>
>
>
> 2011/1/29 Mikel Maron <mikel_maron at yahoo.com>
>
>>  Strategic
>>
>> We've been asked to look into Routing by Nic Roets, author of Gosmore. On
>> the agenda for next time,
>>
>> Best
>> Mikel
>>
>> == Mikel Maron ==
>> +254(0)724899738 @mikel s:mikelmaron
>> http://mapkibera.org/
>> http://wiki.openstreetmap.org/index.php/Haiti
>>
>>
>> ----- Forwarded Message ----
>> *From:* Richard Weait <richard at weait.com>
>> *To:* Mikel Maron <mikel_maron at yahoo.com>
>> *Cc:* Nic Roets <nroets at gmail.com>; Kai Krueger <kakrueger at gmail.com>
>> *Sent:* Sat, January 29, 2011 5:46:36 AM
>> *Subject:* Re: Routing on OSM
>>
>>
>> As I understand it Nic has access to dev and is running a router there.
>>
>> http://nroets.dev.openstreetmap.org/demo/index.html?layers=B000FTFTT
>>
>> SWG involvement might extend as far as "Does OSMF want routing on our
>> hardware?" We looked at that a while back but not since formalizing
>> SWG. We should probably put it on the agenda some time and take a run
>> at it.
>>
>> If OSMF wants to run routing services the following questions likely
>> belong to TWG:
>>
>> - Which routing engines meet our requirements for stability and
>> performance?
>> - What infrastructure is required to run it?
>> - What will be required to manage growth of that service?
>> - How should we implement routing?
>>
>> _______________________________________________
>> Strategic mailing list
>> Strategic at openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/strategic
>>
>>
>
> _______________________________________________
> Strategic mailing list
> Strategic at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/strategic
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/strategic/attachments/20110130/ea512432/attachment-0001.html>