[Talk-GB] OpenKent, OSM coverage estimation

[Richard Fairhurst]

TimSC wrote:
> On 07/06/11 14:37, Richard Fairhurst wrote:
>> You don't need to put stuff into OSM to make it mashable-uppable. Most
>> competent licences will have a Collective Work/Database provision to
>> enable this.
> While this this strictly true it is sometimes hard to associate 
> external records with specific OSM objects. Some importing of reference 
> and ID numbers makes this easier.

It's only hard because no-one's yet built a tool to do it.

You don't have to be that other double-barrelled Tim to understand that
linked data is the coming thing and that (as indeed timbl has pointed out)
OSM is ideally suited to be part of this new world. But you have to have
some way of linking, and stuffing OSM with every single id of every single
dataset that might want to link to it is self-evidently _not_ the way to do
it. 

It isn't as complex as you'd think. You could provide an OSM service which
ensures some degree of id memory. Alternatively, you could provide a way of
fuzzy matching without ids ("the chemist around 52.9346, -1.87639"). There's
huge amounts of prior art to work from (Yahoo WOEIDs and all that).

If only we had more people who were prepared to pull their boots on and
actually do stuff :(

> And back to my original point, I am still not sure if under the new OSM 
> license if I can mash up OSM data with, for example, OGL data as a 
> "produced work".

As a Produced Work, yes, you can - no matter whether OGL is compatible with
ODbL+CT. ODbL allows you to make a Produced Work from a Collective Database.
That's right at the top, in 1.0, in the definition of Produced Work.

>>> [...]
>>> Any what if the government dataset is open and stomps on OSM's attempt?
>> OS OpenData is easily the best free geodata available in the UK and I've
>> just used it (in preference to OSM) to make a lovely paper map, but it
>> hasn't killed OSM yet. :)
> Again, separate issue. Ok, contributors still contribute to OSM but how 
> are we doing on users actually using OSM when it is incomplete compared 
> to other data sets? Would we have more users if our coverage was better? 
> I argue, yes of course.

Well, I can only give you my own view as a map data consumer who's recently
chosen to make a map from OS OpenData rather than OSM, and that is:
completeness isn't the issue.

OSM coverage is very very good in many areas. The OSM community is generally
very responsive to requests like "I'd like to use this area, would anyone
like to map it?" I've done this in the past and, one week later, the map was
complete, surveyed by hand.

The show-stoppers are different. OS OpenData has a more consumer-friendly
licence than OSM. That's huge. Second, OSM is much harder to use, not
principally because of tagging (which is trivial to parse) but because of
vastly varying standards of precision. OS OpenData is both consistently
attributed and generalised to a particular scale.

If I'd wanted to make the West Oxfordshire Green Travel Map from OSM data, I
know that the incomplete areas would have been mapped if I'd asked. (As it
happens, they have been done since, independently.) But completeness wasn't
the problem.

>> In a few cases, manually importing data can indeed be a useful tool. The
>> high-resolution rivers and streams in VectorMap District are quite useful
>> _if_ you know the stream is indeed there, which obviously VMD doesn't
>> tell
>> you.
> You are referencing the common guideline that mappers should only edit 
> areas they have been to. I don't follow that guideline blindly, as you 
> pointed out. Steve Chilton and myself have traced many streams from 
> decades old maps. We like to think we are improving OSM and no one has 
> complained about a specific stream edit yet, as far as I am aware. I had 
> a few (four or five) queries about specific roads but the questions are 
> always requests for confirmation rather than demands to stop importing.
> 
> As far as I understand, your vision of a map which has only direct 
> knowledge and survey would leave many countryside and mountainous areas 
> very bare. You obviously consider this acceptable (and actually that 
> view has some merit). Many tracing contributors don't. A near blank 
> walking map is nearly useless - which is what would result, if we only 
> have map data on OSM contributor accessible places.
> 
> I guess you already thought of all this, so time for me to shut up on 
> that point!

I'm not against tracing areas, or importing individual geometries, where the
mapper has subject knowledge. If you know there's a stream there, by all
means use Bing, or indeed VMD, to get the geometry right. (Indeed, that's
why I collected and scanned NPE, and built the Vector Backgrounds feature
into Potlatch 2, respectively.) And I'm certainly not against iterative
improvement - that's the essence of OSM.

But I am vehemently against contributions, of any sort, where you don't have
knowledge of what you're contributing.

No-one has a _right_ to sit down in front of Potlatch [other editors are
available ;) ] and "contribute" if they don't have anything to contribute.
The "I want to edit" mentality is what's made Wikipedia a fraction of the
site it could be. That's what has killed the usefulness of the OSM wiki for
so many things. The OSM map is too good to let the same happen.

I'm not saying this out of ideology. Whatever works, works. There are
several examples in the UK that mass uninformed tracing doesn't work.

Much of this is another Simple Matter of Programming. We need to make it
easier for people to bring data in manually and carefully, than to blindly
trace or import. We need to make it easier for people to contribute by
annotating map tiles and imagery. I'm doing my bit. Why not help?

(As for countryside and mountainous areas - no way! Have you seen the
Mineral Tramways in Cornwall? The Brecon Beacons, Snowdonia, the Lakes? Our
footpath coverage here in West Oxfordshire? Nick W's work in Hampshire over
many years? OSM's UK countryside coverage is _outstanding_, better than
anything except OS Landranger/Explorer, and it's almost all down to survey.
Can you get surveyors into even the most remote areas of Britain and to
upload their observations? I think Geograph fairly conclusively proves that
you can. ;) )

>> [...]
>>   I'm sure there's been an example
>> where an import has been significant in the success of OSM in the UK but
>> I'm struggling to think of one. Maybe someone else can help?
> It depends on your definition of "import" (obviously). If you include 
> tracing, I traced 90% of SE London and then Semantic Tourist used that 
> in walking papers to survey it personally. Would that be an example? It 
> also fits my vision of "import and improve".

Eeek. Ok, how do I say this without seeming really very rude and
confrontational...

It's basically only helpful if you presume that Semantic Tourist (or
whoever) is incapable of drawing the geometries themselves. I honestly doubt
that. Even if true, your attentions would be better spent improving either
our tools or our documentation so that not just Semantic Tourist, but a
million mappers the world over, are empowered to map for themselves.

Again, trying to remain polite, but I'm afraid I've spoken to so many people
in the OSM London community who have found your "mapping" in South London a
genuine hindrance, that I find it difficult to see any value in it for the
community. I don't like to be quite so unequivocally dismissive of someone's
work but it does appear to be a very common experience among London mappers.

And yes, I am particularly concerned in connection with the relicensing.
Tracing blindly from Yahoo would not attract copyright protection in many
jurisdictions (e.g. the US); it's unclear in the UK. Nonetheless it is, in
my eyes at least, troublesome that someone can do the trivial work of
tracing from aerial imagery for v1 of a way (possibly just a few seconds'
work, a small number of nodes, no metadata beyond highway=residential); that
an on-the-ground mapper can come along and spend an hour surveying and
adding metadata, better geometry and so on; and yet the v1 trivial tracer,
by claiming copyright where none may exist, has a right of ransom over the
hour the on-the-ground mapper has put into it.

We are seeing, right now, how this can play out. This exact thing (albeit
import rather than tracing) has happened in Ottawa, where an irresponsible
importer has decided, without any foundation, to remove their trivial
contribution using ODbL+CT as a pretext. Hours of on-the-ground mapping has
been lost as a result.

Fortunately this case is being reverted and I'm not suggesting that you
would do anything so irresponsible. But it's been very interesting to see
the community reaction. People on talk-ca are, quite rightly, furious that
someone has taken it on themselves to delete the community's work. I think
the situation would be the same in the UK if anyone were to unreasonably
withhold permission to relicense their trivial contribution. I hope it
doesn't come to that.

> [...]
> It would be hard to argue that OS Opendata has not been an asset to 
> OSM...? There are many country roads that are not sign posted but are in 
> OS Locator - and now in OSM.

OS OpenData is certainly useful in some circumstances but that's not
especially one of them. The quality of naming in OS Locator is surprisingly
poor (I've posted here before about its several mistakes in little old
Charlbury). The rural road names look seductively helpful but aren't: IMX
very often no-one actually calls the roads by those names, whereas at other
times locals will use a name not recorded on Locator. I haven't copied
Locator names for rural roads round Charlbury for precisely that reason,
even if it does screw up our OSM Analysis stats - it's not so much "name" or
"not:name" as "might-be-name". ;)

Locator is principally useful for flagging up "go and survey this". Again,
there is a big difference between tracing/importing areas of which you have
subject knowledge, and blindly tracing/importing areas of which you don't.
If you don't have subject knowledge, it doesn't work. 

> [...]
> If you hope that relicensing will reduce the amount of OS Opendata, I 
> think you might be disappointed. Many significant contributors have 
> signed the CTs on the assumption that LWG will find a way to allow 
> continued OS Opendata importing. [1][2]

Yet there does, rather curiously, appear to be a correlation between
importers/tracers and ODbL refuseniks/not-yet-confirmedniks. I don't know
why that is, but I do suspect that the result will be a little less
importing, and a bit more mapping, when the switch happens. 

cheers
Richard



--
View this message in context: http://gis.638310.n2.nabble.com/OpenKent-OSM-coverage-estimation-tp6434777p6454031.html
Sent from the Great Britain mailing list archive at Nabble.com.