Re: [OSM-talk] Nominatim on the main page

[Sarah Hoffmann]

On Mon, Feb 19, 2018 at 10:59:31AM +0100, Maarten Deen wrote:
> On 2018-02-19 10:17, Sarah Hoffmann wrote:
> > On Sun, Feb 18, 2018 at 08:12:45PM +0100, Maarten Deen wrote:
> > > On 2018-02-18 20:07, Paul Johnson wrote:
> > > > On Sun, Feb 18, 2018 at 12:53 PM, Maarten Deen 
> > > > wrote:
> > > >
> > > > > On 2018-02-18 19:28, Tom Hughes wrote:
> > > 
> > > > > I can't comment about how the algorithm works because I don't know
> > > > > anything about it. I'm just saying that we do tell it the viewbox
> > > >
> > > >  It appears to me that the bounding box is used when searching places
> > > > (towns, cities) or streets, but not when searching objects like shops
> > > > or restaurants.
> > > > For instance, searching for a McDonald's always gives me the
> > > > McDonald's at 1351, George Dieter Drive, El Paso City, El Paso County,
> > > > Texas, 79936, Verenigde Staten van Amerika
> > 
> > To fix that please delete all the wikipedia=McDonalds tags from
> > the McDonalds restaurants that show up inappropriately. Nominatim uses
> > the wikipedia links to determine how well known a place might be and
> > ranks places with a wikipedia tag higher.
> 
> I would expect that a bounding box has precedence over other tags. Why would
> a wikipedia tag have precedence over the bounding box and a name tag not?

It is not a question of precedence. Nominatim looks at different
factors at the same time: the view box, how well-known a place
is (aka wikipedia importance), how well the name of the place matches
your query etc. It takes all these into account weighs them against
each other and comes to a ranking of results.

> But then I still don't understand.
> In the bugreport I have the example of the shop "kruidvat" (it is a chain of
> stores in the Netherlands).
> The bounding box is 6.16575,51.36926,6.17049,51.36759 which centers on the
> Kruidvat store in Venlo [1]. Nominatim returns the kruidvat in Amsterdam
> [2].
> Both nodes have a website tag with the same value. Both nodes have the same
> tags, expect that one has a source tag.
> Then still, why is the boundingbox not looked at _at all_? It's not like
> it's the second or third result, it is the 12th result where all other
> results have similar tags and the results are the same whatever bounding box
> you use.

Interesting. So in this case the importance actually happend to accidentally
cancel out the viewbox influence. I've pushed a preliminary fix to the osm.org
instance. It won't fix the McDonalds or Walmart issues though. They are problems
of a different kind.

Sarah

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Jorge Gustavo Rocha]

Thanks, Sarah! That really helps to start contributing.

Regards,

Gustavo

On 19-02-2018 09:31, Sarah Hoffmann wrote:
> On Sun, Feb 18, 2018 at 08:15:57PM +, Jorge Gustavo Rocha wrote:
>>
>>
>> On 18-02-2018 19:21, Milo van der Linden wrote:
>>> With 103 open issues and 12 open pull requests, I would love to
>>> volunteer to at least help get those cleared first. Given the (very
>>> positive, I am glad so many people are acting on this thread)
>>> activity, I think if everybody lends a couple of hours of code this
>>> week we can get nominatim ready to make some progress.
>>>
>>
>> +1
>>
>> I did not blame before, because I never contributed to nominatim. I'll
>> take some time this week to review issues and PR (although this week is
>> the QGIS hackfest).
>>
>> But definitely, I'll use the open data day [1] dedicated to nominatim.
>> Maybe we can have a virtual meeting on March 3rd dedicated to nominatim
>> to coordinate actions.
> 
> Awesome. Nominatim can use all the help it can get.
> 
> The open PRs marked 'Changes requested' are an excellent place to start
> with. Those are changes that are good in general but unfortunately have
> been abandoned by the original author very close before the finish line.
> 
> I have also marked a couple of issue as 'simple'. They are good to get
> used to the code base before tackling the harder ones.
> 
> Kind regards
> 
> Sarah
> 
> ___
> talk mailing list
> talk@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk
> 

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Maarten Deen]

On 2018-02-19 10:17, Sarah Hoffmann wrote:

On Sun, Feb 18, 2018 at 08:12:45PM +0100, Maarten Deen wrote:

On 2018-02-18 20:07, Paul Johnson wrote:
> On Sun, Feb 18, 2018 at 12:53 PM, Maarten Deen 
> wrote:
>
> > On 2018-02-18 19:28, Tom Hughes wrote:

> > I can't comment about how the algorithm works because I don't know
> > anything about it. I'm just saying that we do tell it the viewbox
>
>  It appears to me that the bounding box is used when searching places
> (towns, cities) or streets, but not when searching objects like shops
> or restaurants.
> For instance, searching for a McDonald's always gives me the
> McDonald's at 1351, George Dieter Drive, El Paso City, El Paso County,
> Texas, 79936, Verenigde Staten van Amerika


To fix that please delete all the wikipedia=McDonalds tags from
the McDonalds restaurants that show up inappropriately. Nominatim uses
the wikipedia links to determine how well known a place might be and
ranks places with a wikipedia tag higher.


I would expect that a bounding box has precedence over other tags. Why 
would a wikipedia tag have precedence over the bounding box and a name 
tag not?


But then I still don't understand.
In the bugreport I have the example of the shop "kruidvat" (it is a 
chain of stores in the Netherlands).
The bounding box is 6.16575,51.36926,6.17049,51.36759 which centers on 
the Kruidvat store in Venlo [1]. Nominatim returns the kruidvat in 
Amsterdam [2].
Both nodes have a website tag with the same value. Both nodes have the 
same tags, expect that one has a source tag.
Then still, why is the boundingbox not looked at _at all_? It's not like 
it's the second or third result, it is the 12th result where all other 
results have similar tags and the results are the same whatever bounding 
box you use.


I can accept that the website tag raises the ranking, but that doesn't 
apply in this case since all nodes have the same website tag. It still 
looks like the bounding box is not used at all.


I have another example: [3] centers on a restaurant called "Sint Maria" 
[4]
Entering this in Nominatim [5] gives tram stops in Brussels, churches 
and graveyards in Germany. None of which have a website tag.


[1] http://www.openstreetmap.org/node/2931123803
[2] http://www.openstreetmap.org/node/2862337526
[3] http://www.openstreetmap.org/#map=19/51.32897/5.97973
[4] http://www.openstreetmap.org/node/3008677405
[5] 
https://nominatim.openstreetmap.org/search.php?q=sint+maria_geojson=1=5.97762%2C51.32981%2C5.98209%2C51.32813



Regards,
Maarten

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Sarah Hoffmann]

On Sun, Feb 18, 2018 at 08:15:57PM +, Jorge Gustavo Rocha wrote:
> 
> 
> On 18-02-2018 19:21, Milo van der Linden wrote:
> > With 103 open issues and 12 open pull requests, I would love to
> > volunteer to at least help get those cleared first. Given the (very
> > positive, I am glad so many people are acting on this thread)
> > activity, I think if everybody lends a couple of hours of code this
> > week we can get nominatim ready to make some progress.
> > 
> 
> +1
> 
> I did not blame before, because I never contributed to nominatim. I'll
> take some time this week to review issues and PR (although this week is
> the QGIS hackfest).
> 
> But definitely, I'll use the open data day [1] dedicated to nominatim.
> Maybe we can have a virtual meeting on March 3rd dedicated to nominatim
> to coordinate actions.

Awesome. Nominatim can use all the help it can get.

The open PRs marked 'Changes requested' are an excellent place to start
with. Those are changes that are good in general but unfortunately have
been abandoned by the original author very close before the finish line.

I have also marked a couple of issue as 'simple'. They are good to get
used to the code base before tackling the harder ones.

Kind regards

Sarah

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Sarah Hoffmann]

On Sun, Feb 18, 2018 at 08:12:45PM +0100, Maarten Deen wrote:
> On 2018-02-18 20:07, Paul Johnson wrote:
> > On Sun, Feb 18, 2018 at 12:53 PM, Maarten Deen 
> > wrote:
> > 
> > > On 2018-02-18 19:28, Tom Hughes wrote:
> 
> > > I can't comment about how the algorithm works because I don't know
> > > anything about it. I'm just saying that we do tell it the viewbox
> > 
> >  It appears to me that the bounding box is used when searching places
> > (towns, cities) or streets, but not when searching objects like shops
> > or restaurants.
> > For instance, searching for a McDonald's always gives me the
> > McDonald's at 1351, George Dieter Drive, El Paso City, El Paso County,
> > Texas, 79936, Verenigde Staten van Amerika

To fix that please delete all the wikipedia=McDonalds tags from
the McDonalds restaurants that show up inappropriately. Nominatim uses
the wikipedia links to determine how well known a place might be and
ranks places with a wikipedia tag higher. That naturally only works
when the wikipedia tags actually link to a wikipedia page that
describes the object. It leads to funny results when the link goes
to category pages or, like in this case, to the company description.

Alternatively: I've proposed a GSoC to overhaul the Wikipedia
importances that Nominatim uses. Getting rid of this particular
problem from the Nominatim side would be part of this job.

For more information, see:
https://wiki.openstreetmap.org/wiki/Google_Summer_of_Code/2018/Project_Ideas#Nominatim

There are also two other topics proposed and if you have another particular
itch you want to sratch, there are surely ways they can be transformed into
a GSoC topic. Just send me a email or open an issue in github. It would be 
wonderful, if
we find some students interested in geocoding this year.

Kind regards

Sarah

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Komяpa]

>
> Have a look at the OSMF board, a mixed bunch of people elected by the
> members. Are you sure that a seasoned developer or sysadmin would even
> *want* a paid job where they are subject to the whims of an elected
> board, with a potentially modified "strategy" year after year (as
> majorities change due to new elections)? Would that not be a job like
> Dilbert's with his pointy-haired boss?
>

Would such a person take this position voluntarily and unpaid?

Existing stable devs and sysadmins are paid with "unlimited power" in the
project now. Is that a good thing? If we get more devs/sysadmins "for
free", we're splitting the "unlimited" into halves effectively making it
limited and much less appealing. Can we replace "a free unlimited power"
payment we give to dev/sysadmins now with any other scalable means, be that
money or anything other?
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page / Serges blog post

[Simon Poole]

Maybe it would be a good idea for everybody to take a deep breath and
count to 10?

For those relatively new to the project you should know that Serge has
been going on about these "issues" for years, some since I first had the
pleasure of  meeting him. That doesn't  mean that his points are more or
less valid, but the general sense of urgency due to a click-bait title
stuck above them is rather misplaced.

As to the points themselves, they are a mix of things from latent issues
that could easily turn serious (lack of man-power in some core areas),
over Serges pet peeves (nominatim doesn't do US junctions), to
trade-offs that he might disagree with, but are nonetheless not without
merit (for example layers, that now and then would be nice to have, but
are rather at odds with OSMs more holistic approach to modelling).

For now I just want to touch on the money side of things and employing
staff. The amounts available for speculative financial investment
naturally have nothing to do with operating a self-sustaining
(geo-services-) business. And if anybody doubts the difficulties of
doing the later, we just had the likely best ever funded geo-startup
that was literally able to throw money out of the window, miserably fail
when it was supposed to stand on its own feet and actually turn
geo-coding, map tiles etc, "over a free tier" in to a sustainable business.

Now the OSMF might have a slight advantage due to brand recognition, but
that doesn't change the fact that there is not very much money available
out there for such services and that such a pivot would come at the cost
of ruining the many legit small business that are small but sustainable,
and of which many contribute in one way or another to OSM software
development and operations.

All that said, OSM operations and core software development are
dependent on a very very small number of people, some of which have been
with us since the project started, we really need more hands full stop.
If paying for staff is the solution, difficult to say, what is clear is
that we would need a substantially larger continuous stream of donations
(think a pineapple fund sized donation every quarter) for any meaningful
number of operational staff.

In any case I'll be asking for help in a follow up  posting on a
specific ops issue and expect to see lots of people volunteering :-).

Simon

PS: The limitation in Nominatims POI search is discussed roughly every
half year, well documented and understood, it would undoubtedly be nice
if it was fixed and if it is only to avoid the continuous  surprise
re-discovery of it.



signature.asc
Description: OpenPGP digital signature
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Frederik Ramm]

Jason,

On 02/18/2018 08:07 PM, Jason Remillard wrote:
> There is plenty of money around this space to pay for a full time
> system administrator staff and some developers. Pokémon Go netted 600
> million dollars in the first three months. Mapbox just go $164 million
> dollar investment. I don't understand why you, Tom, lonvia are not
> paid, full time employees of OSMF by now. Mapbox is doing a great job
> with ID development, but obviously they are not going to seriously
> fund our tiles and geocoding.

Many people seem to believe that the natural growth path for a nonprofit
organisation is identical to that of a for-profit organisation. Acquire
more funds, hire more people, invest, grow, rinse & repeat - bigger,
better, faster, more, year after year.

This is certainly a *possible* course of action.

But it is desirable?

Have a look at the OSMF board, a mixed bunch of people elected by the
members. Are you sure that a seasoned developer or sysadmin would even
*want* a paid job where they are subject to the whims of an elected
board, with a potentially modified "strategy" year after year (as
majorities change due to new elections)? Would that not be a job like
Dilbert's with his pointy-haired boss?

Ah, you'll say, easy: We hire a CEO in addition to that, so there is
more continuity, and the CEO can then boss the technicians around, and
the board only adjusts the general direction occasionally. That's also
something people often suggest, again following the usual corporate lines.

Growth of this kind entrenches the power of the established organisation
over the volunteers. Inevitably, the organisation shifts from the
initial "a bunch of volunteers have founded an organisation and elected
a board to do the basic housekeeping of financials and intellectual
property rights" to "a central organisation manages their volunteers".

The growth of the central organisation brings with it increased funding
requirements (salary for CEO, developers, and sysadmins; soon after,
cost for a fixed office, office management, community managers, and so
on). This money has to come from somewhere (add salary for professional
fundraisers). Whoever gives us the money can make all sorts of demands
on our organisation and community. Shrinking the organisation is rarely
an option, so we'll need the funds to continue to live, and we are much
easier to control by "big capital" (if you look at usual funding
sources, you can safely say "big US capital").

We will be subject to politics much more than now; we will have to
publish regular success stories (and gloss over failures) just like any
business does to maintain a good image. We will have to kick out people
in our ranks who soil our image lest the funders will back out. We'll
have to start doing silly side projects that the community would never
have started in their own, just to tick some boxes with a funding partner.

Do we want to sell our soul? And for what precisely in return?

These are important questions to think about. I am sure there are
potential growth paths that do *not* require us to blindly submit to the
ways of commerce - just like we never blindly submitted to traditional
GIS practices ;)

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frede...@remote.org  ##  N49°00'09" E008°23'33"

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Jorge Gustavo Rocha]

On 18-02-2018 19:21, Milo van der Linden wrote:
> With 103 open issues and 12 open pull requests, I would love to
> volunteer to at least help get those cleared first. Given the (very
> positive, I am glad so many people are acting on this thread)
> activity, I think if everybody lends a couple of hours of code this
> week we can get nominatim ready to make some progress.
> 

+1

I did not blame before, because I never contributed to nominatim. I'll
take some time this week to review issues and PR (although this week is
the QGIS hackfest).

But definitely, I'll use the open data day [1] dedicated to nominatim.
Maybe we can have a virtual meeting on March 3rd dedicated to nominatim
to coordinate actions.

[1] March 3rd, 2018, http://opendataday.org

> 2018-02-18 20:12 GMT+01:00 Maarten Deen :
>> On 2018-02-18 20:07, Paul Johnson wrote:
>>>
>>> On Sun, Feb 18, 2018 at 12:53 PM, Maarten Deen 
>>> wrote:
>>>
 On 2018-02-18 19:28, Tom Hughes wrote:
>>
>>
 I can't comment about how the algorithm works because I don't know
 anything about it. I'm just saying that we do tell it the viewbox
>>>
>>>
>>>  It appears to me that the bounding box is used when searching places
>>> (towns, cities) or streets, but not when searching objects like shops
>>> or restaurants.
>>> For instance, searching for a McDonald's always gives me the
>>> McDonald's at 1351, George Dieter Drive, El Paso City, El Paso County,
>>> Texas, 79936, Verenigde Staten van Amerika
>>>
>>>  Keen ask.  I think a _lot_ of the complaints regarding Nominatim that
>>> have come up could be fixed by using the bounding box context for all
>>> items.  I may have incorrectly opened an Osmand bug [1] on the
>>> assumption Nominatim was consistently using the geocontext for all
>>> results.
>>
>>
>> I've opened a Nominatim bug for it.
>> https://github.com/openstreetmap/Nominatim/issues/930
>>
>>
>> Maarten
>>
>> ___
>> talk mailing list
>> talk@openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/talk
> 
> 
> 

-- 
OSGeo Portugal
NIF: 509588190
email: ge...@osgeopt.pt
https://www.osgeopt.pt/

Próximo evento: SASIG 2017, 20 a 22 de novembro, Porto
Mais informações: http://osgeopt.pt/sasig2017/

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Grant Slater]

On 18 February 2018 at 19:07, Jason Remillard  wrote:
> Hi Grant,
>
> If you read Serge's post, he is quite clear on what his preferred
> solution to this resourcing problem. Our tiles and Nominatim services
> should have terms of services that include paid higher levels that
> support the foundation, which in turn pays for the infrastructure.
> When somebody is using too many tiles, or using too much of Nominatim
> resources, or wants Nominatim to work better, etc, before chasing them
> away, the foundation should ask them for money to support the
> requested services.
>

OpenStreetMap.org services are abused precisely because the are 1)
Free to use, 2) they are open to use (no keys etc) and 3) for most
part we don't actually limits the usage of the services (There isn't
actually a 5% cap on tiles).
I'd feel uncomfortable with an Open Data project charging for services.

I'm a reasonably good sysadmin, but people smarter than me need to
decide on the direction of the project. ;-)

Interesting parallel discussion thread going on on twitter, maybe OSMF
should be "willing to pay a developer a market-rate salary to own this
problem" using some of the Pineapple Fund donation:
Jumping off point: https://twitter.com/richardf/status/965301288933175296

I'd be all for that.

> There is plenty of money around this space to pay for a full time
> system administrator staff and some developers. Pokémon Go netted 600
> million dollars in the first three months. Mapbox just go $164 million
> dollar investment. I don't understand why you, Tom, lonvia are not
> paid, full time employees of OSMF by now. Mapbox is doing a great job
> with ID development, but obviously they are not going to seriously
> fund our tiles and geocoding.
>

Small steps and iterating sounds like a healthier option to me.

> There is a great need for what OSM does, we just need to ask for
> money, rather than acting like a charity, begging for handouts.
>

Let's Encrypt is a "charity" and is "begging for handouts", but is
clearly winning and has massively moved the internet forward in the
last 2 years. I don't think OpenStreetMap needs to become a commercial
project to survive.

Kind regards,

Grant

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Milo van der Linden]

With 103 open issues and 12 open pull requests, I would love to
volunteer to at least help get those cleared first. Given the (very
positive, I am glad so many people are acting on this thread)
activity, I think if everybody lends a couple of hours of code this
week we can get nominatim ready to make some progress.

2018-02-18 20:12 GMT+01:00 Maarten Deen :
> On 2018-02-18 20:07, Paul Johnson wrote:
>>
>> On Sun, Feb 18, 2018 at 12:53 PM, Maarten Deen 
>> wrote:
>>
>>> On 2018-02-18 19:28, Tom Hughes wrote:
>
>
>>> I can't comment about how the algorithm works because I don't know
>>> anything about it. I'm just saying that we do tell it the viewbox
>>
>>
>>  It appears to me that the bounding box is used when searching places
>> (towns, cities) or streets, but not when searching objects like shops
>> or restaurants.
>> For instance, searching for a McDonald's always gives me the
>> McDonald's at 1351, George Dieter Drive, El Paso City, El Paso County,
>> Texas, 79936, Verenigde Staten van Amerika
>>
>>  Keen ask.  I think a _lot_ of the complaints regarding Nominatim that
>> have come up could be fixed by using the bounding box context for all
>> items.  I may have incorrectly opened an Osmand bug [1] on the
>> assumption Nominatim was consistently using the geocontext for all
>> results.
>
>
> I've opened a Nominatim bug for it.
> https://github.com/openstreetmap/Nominatim/issues/930
>
>
> Maarten
>
> ___
> talk mailing list
> talk@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk



-- 
Milo van der Linden
web: dogodigi
tel: +31-6-16598808

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Maarten Deen]

On 2018-02-18 20:07, Paul Johnson wrote:

On Sun, Feb 18, 2018 at 12:53 PM, Maarten Deen 
wrote:


On 2018-02-18 19:28, Tom Hughes wrote:



I can't comment about how the algorithm works because I don't know
anything about it. I'm just saying that we do tell it the viewbox


 It appears to me that the bounding box is used when searching places
(towns, cities) or streets, but not when searching objects like shops
or restaurants.
For instance, searching for a McDonald's always gives me the
McDonald's at 1351, George Dieter Drive, El Paso City, El Paso County,
Texas, 79936, Verenigde Staten van Amerika

 Keen ask.  I think a _lot_ of the complaints regarding Nominatim that
have come up could be fixed by using the bounding box context for all
items.  I may have incorrectly opened an Osmand bug [1] on the
assumption Nominatim was consistently using the geocontext for all
results.


I've opened a Nominatim bug for it.
https://github.com/openstreetmap/Nominatim/issues/930

Maarten

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Jason Remillard]

Hi Grant,

If you read Serge's post, he is quite clear on what his preferred
solution to this resourcing problem. Our tiles and Nominatim services
should have terms of services that include paid higher levels that
support the foundation, which in turn pays for the infrastructure.
When somebody is using too many tiles, or using too much of Nominatim
resources, or wants Nominatim to work better, etc, before chasing them
away, the foundation should ask them for money to support the
requested services.

There is plenty of money around this space to pay for a full time
system administrator staff and some developers. Pokémon Go netted 600
million dollars in the first three months. Mapbox just go $164 million
dollar investment. I don't understand why you, Tom, lonvia are not
paid, full time employees of OSMF by now. Mapbox is doing a great job
with ID development, but obviously they are not going to seriously
fund our tiles and geocoding.

There is a great need for what OSM does, we just need to ask for
money, rather than acting like a charity, begging for handouts.

Jason


> Nominatim effectively only has 1 maintainer / primary contributor:
> https://github.com/openstreetmap/Nominatim/graphs/contributors
> lonvia spends a significant amount of her volunteered time defending
> the service again abusers.
> https://operations.osmfoundation.org/policies/nominatim/
> Single users can overload nominatim by sending 1000s of automated
> requests per second, the same holds true for the tile servers we run:
> https://operations.osmfoundation.org/policies/tiles/
>
> Nearly as bad contribution wise is the server infrastructure code:
> https://github.com/openstreetmap/chef/graphs/contributors
>
> What we badly need is more help from developers and chef operations people.
>
> How about someone in the community organises a Beauty Parade / Cross
> Comparison of different open OpenStreetMap GeoCoders?
>
> The OpenStreetMap Operations team prefers to be able to self-host the
> infrastructure required to run OpenStreetMap.org. Reasons: Privacy /
> Maintainability / Availability. We don't go this route for a
> navigation / router as they seem to be under constant flux, without a
> clear "winner" and system their requirements keep climbing.
>
> Kind regards,
>
> Grant
>
> ___
> talk mailing list
> talk@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Paul Johnson]

On Sun, Feb 18, 2018 at 12:53 PM, Maarten Deen  wrote:

> On 2018-02-18 19:28, Tom Hughes wrote:
>
>> On 18/02/18 18:04, Paul Johnson wrote:
>>
>>   OK, so what can we do about this problem?  For example, go to Jenks,
>>> Oklahoma.  Search for Walmart.  First result isn't the Walmart Neighborhood
>>> Market across the street from the Riverside Market shopping center.  It's
>>> not even the Walmart Supercenter across from Oral Roberts University.  Or
>>> even the Walmart Neighborhood Market further from downtown Jenks, but still
>>> in Jenks, over by Haddington Heights.  The results instead are in Sunset
>>> Harbor, Florida; Saint Louis, Missouri; Tallahassee, Florida, El Paso,
>>> Texas; and Fairbanks, Alaska.
>>>
>>> I get that if you get specific and actually type "Walmart jenks" in, you
>>> get the one in Haddington Heights, and then the one in Glenpool, Oklahoma
>>> (slightly odd for that one, but not entirely unreasonable).  But this isn't
>>> what most people are going to do.  I also get that the homepage isn't
>>> /meant/ to be used as the product itself, just a demonstration, but we
>>> really should be putting our best foot forward there.
>>>
>>
>> I can't comment about how the algorithm works because I don't know
>> anything about it. I'm just saying that we do tell it the viewbox
>>
>
> It appears to me that the bounding box is used when searching places
> (towns, cities) or streets, but not when searching objects like shops or
> restaurants.
> For instance, searching for a McDonald's always gives me the McDonald's at
> 1351, George Dieter Drive, El Paso City, El Paso County, Texas, 79936,
> Verenigde Staten van Amerika


 Keen ask.  I think a *lot* of the complaints regarding Nominatim that have
come up could be fixed by using the bounding box context for all items.  I
may have incorrectly opened an Osmand bug
 on the assumption
Nominatim was consistently using the geocontext for all results.
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Maarten Deen]

On 2018-02-18 19:28, Tom Hughes wrote:

On 18/02/18 18:04, Paul Johnson wrote:

  OK, so what can we do about this problem?  For example, go to Jenks, 
Oklahoma.  Search for Walmart.  First result isn't the Walmart 
Neighborhood Market across the street from the Riverside Market 
shopping center.  It's not even the Walmart Supercenter across from 
Oral Roberts University.  Or even the Walmart Neighborhood Market 
further from downtown Jenks, but still in Jenks, over by Haddington 
Heights.  The results instead are in Sunset Harbor, Florida; Saint 
Louis, Missouri; Tallahassee, Florida, El Paso, Texas; and Fairbanks, 
Alaska.


I get that if you get specific and actually type "Walmart jenks" in, 
you get the one in Haddington Heights, and then the one in Glenpool, 
Oklahoma (slightly odd for that one, but not entirely unreasonable).  
But this isn't what most people are going to do.  I also get that the 
homepage isn't /meant/ to be used as the product itself, just a 
demonstration, but we really should be putting our best foot forward 
there.


I can't comment about how the algorithm works because I don't know
anything about it. I'm just saying that we do tell it the viewbox


It appears to me that the bounding box is used when searching places 
(towns, cities) or streets, but not when searching objects like shops or 
restaurants.
For instance, searching for a McDonald's always gives me the McDonald's 
at 1351, George Dieter Drive, El Paso City, El Paso County, Texas, 
79936, Verenigde Staten van Amerika


Maarten

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Tom Hughes]

On 18/02/18 18:40, Paul Johnson wrote:
On Sun, Feb 18, 2018 at 12:31 PM, Tom Hughes > wrote:


On 18/02/18 18:05, Paul Johnson wrote:

On Sun, Feb 18, 2018 at 11:58 AM, Grant Slater

>> wrote:

     How about someone in the community organises a Beauty
Parade / Cross
     Comparison of different open OpenStreetMap GeoCoders?


How about search against multiple geocoders instead of just
Nominatim and GeoNames?


We have all the infrastructure to do that if you have suggestions
about other geocoders we could use.


Not having tried it myself, I have heard positive things on Reddit and 
Telegram about Photon .


Sorry I probably wasn't very clear. I meant an actual service we could
direct queries to, not just come code.

I mean we probably could come up with hardware to run something else
but it would need somebody prepared to do the work of running it.

Tom

--
Tom Hughes (t...@compton.nu)
http://compton.nu/

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Paul Johnson]

On Sun, Feb 18, 2018 at 12:31 PM, Tom Hughes  wrote:

> On 18/02/18 18:05, Paul Johnson wrote:
>
>> On Sun, Feb 18, 2018 at 11:58 AM, Grant Slater <
>> openstreet...@firefishy.com > wrote:
>>
>> How about someone in the community organises a Beauty Parade / Cross
>> Comparison of different open OpenStreetMap GeoCoders?
>>
>>
>> How about search against multiple geocoders instead of just Nominatim and
>> GeoNames?
>>
>
> We have all the infrastructure to do that if you have suggestions
> about other geocoders we could use.


Not having tried it myself, I have heard positive things on Reddit and
Telegram about Photon .
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Tom Hughes]

On 18/02/18 18:05, Paul Johnson wrote:
On Sun, Feb 18, 2018 at 11:58 AM, Grant Slater 
> wrote:


How about someone in the community organises a Beauty Parade / Cross
Comparison of different open OpenStreetMap GeoCoders?


How about search against multiple geocoders instead of just Nominatim 
and GeoNames?


We have all the infrastructure to do that if you have suggestions
about other geocoders we could use.

Tom

--
Tom Hughes (t...@compton.nu)
http://compton.nu/

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Tom Hughes]

On 18/02/18 18:04, Paul Johnson wrote:

  OK, so what can we do about this problem?  For example, go to Jenks, 
Oklahoma.  Search for Walmart.  First result isn't the Walmart 
Neighborhood Market across the street from the Riverside Market shopping 
center.  It's not even the Walmart Supercenter across from Oral Roberts 
University.  Or even the Walmart Neighborhood Market further from 
downtown Jenks, but still in Jenks, over by Haddington Heights.  The 
results instead are in Sunset Harbor, Florida; Saint Louis, Missouri; 
Tallahassee, Florida, El Paso, Texas; and Fairbanks, Alaska.


I get that if you get specific and actually type "Walmart jenks" in, you 
get the one in Haddington Heights, and then the one in Glenpool, 
Oklahoma (slightly odd for that one, but not entirely unreasonable).  
But this isn't what most people are going to do.  I also get that the 
homepage isn't /meant/ to be used as the product itself, just a 
demonstration, but we really should be putting our best foot forward there.


I can't comment about how the algorithm works because I don't know
anything about it. I'm just saying that we do tell it the viewbox

Tom

--
Tom Hughes (t...@compton.nu)
http://compton.nu/

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Grant Slater]

On 18 February 2018 at 18:05, Paul Johnson  wrote:
> On Sun, Feb 18, 2018 at 11:58 AM, Grant Slater 
> wrote:
>>
>> How about someone in the community organises a Beauty Parade / Cross
>> Comparison of different open OpenStreetMap GeoCoders?
>
>
> How about search against multiple geocoders instead of just Nominatim and
> GeoNames?

Sure. Which one, hence the Beauty Parade / Cross Comparison? Who will
write the integration code, UI and submit PRs? ;-)
Does OpenStreetMap.org need to host an instance or can we rely on the
3rd party who hosts it? (Privacy / Stability etc)

Kind regards,

Grant

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Paul Johnson]

On Sun, Feb 18, 2018 at 11:58 AM, Grant Slater 
wrote:
>
> How about someone in the community organises a Beauty Parade / Cross
> Comparison of different open OpenStreetMap GeoCoders?
>

How about search against multiple geocoders instead of just Nominatim and
GeoNames?
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Paul Johnson]

On Sun, Feb 18, 2018 at 11:45 AM, Tom Hughes  wrote:

> On 18/02/18 17:34, Paul Johnson wrote:
>
> On the OSM homepage, can we use the visible area (or maybe that plus an
>> exploded offset to a larger surrounding bounds) as a bonding box to be
>> passed to Nominatim for some context when searching?  This animation really
>> drives the problem home.
>>
>
> We already do exactly that.
>
> Sarah can clarify the details but we pass the bounding box to Nominatim
> and I believe it expands that by a factor of two and then prefers any
> results in that area over those further afield.


 OK, so what can we do about this problem?  For example, go to Jenks,
Oklahoma.  Search for Walmart.  First result isn't the Walmart Neighborhood
Market across the street from the Riverside Market shopping center.  It's
not even the Walmart Supercenter across from Oral Roberts University.  Or
even the Walmart Neighborhood Market further from downtown Jenks, but still
in Jenks, over by Haddington Heights.  The results instead are in Sunset
Harbor, Florida; Saint Louis, Missouri; Tallahassee, Florida, El Paso,
Texas; and Fairbanks, Alaska.

I get that if you get specific and actually type "Walmart jenks" in, you
get the one in Haddington Heights, and then the one in Glenpool, Oklahoma
(slightly odd for that one, but not entirely unreasonable).  But this isn't
what most people are going to do.  I also get that the homepage isn't
*meant* to be used as the product itself, just a demonstration, but we
really should be putting our best foot forward there.
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Grant Slater]

On 18 February 2018 at 17:34, Paul Johnson  wrote:
> In the diatribe (no disrespect meant, since it did bring up some relevant
> points that we can use constructively, simply a description of the tone)
> that emacsen posted last week that we've surely all read by now, he does
> bring up one point that I think we can work on for fairly immediate
> improvement.
>
> On the OSM homepage, can we use the visible area (or maybe that plus an
> exploded offset to a larger surrounding bounds) as a bonding box to be
> passed to Nominatim for some context when searching?  This animation really
> drives the problem home.
>
> https://gfycat.com/NiceImperturbableCatfish
>

Not great. :-(

Nominatim effectively only has 1 maintainer / primary contributor:
https://github.com/openstreetmap/Nominatim/graphs/contributors
lonvia spends a significant amount of her volunteered time defending
the service again abusers.
https://operations.osmfoundation.org/policies/nominatim/
Single users can overload nominatim by sending 1000s of automated
requests per second, the same holds true for the tile servers we run:
https://operations.osmfoundation.org/policies/tiles/

Nearly as bad contribution wise is the server infrastructure code:
https://github.com/openstreetmap/chef/graphs/contributors

What we badly need is more help from developers and chef operations people.

How about someone in the community organises a Beauty Parade / Cross
Comparison of different open OpenStreetMap GeoCoders?

The OpenStreetMap Operations team prefers to be able to self-host the
infrastructure required to run OpenStreetMap.org. Reasons: Privacy /
Maintainability / Availability. We don't go this route for a
navigation / router as they seem to be under constant flux, without a
clear "winner" and system their requirements keep climbing.

Kind regards,

Grant

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Tom Hughes]

On 18/02/18 17:45, Tom Hughes wrote:

On 18/02/18 17:34, Paul Johnson wrote:

On the OSM homepage, can we use the visible area (or maybe that plus 
an exploded offset to a larger surrounding bounds) as a bonding box to 
be passed to Nominatim for some context when searching?  This 
animation really drives the problem home.


We already do exactly that.

Sarah can clarify the details but we pass the bounding box to Nominatim
and I believe it expands that by a factor of two and then prefers any
results in that area over those further afield.


Incidentally if you use https://nominatim.openstreetmap.org/ you can
experiment with the affect of the viewbox parameter that we pass by
checking and unchecking the "apply viewbox" checkbox beside the search
button and zooming moving the map to choose a box.

Tom

--
Tom Hughes (t...@compton.nu)
http://compton.nu/

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] Nominatim on the main page

[Tom Hughes]

On 18/02/18 17:34, Paul Johnson wrote:

On the OSM homepage, can we use the visible area (or maybe that plus an 
exploded offset to a larger surrounding bounds) as a bonding box to be 
passed to Nominatim for some context when searching?  This animation 
really drives the problem home.


We already do exactly that.

Sarah can clarify the details but we pass the bounding box to Nominatim
and I believe it expands that by a factor of two and then prefers any
results in that area over those further afield.

Here's the relevant code:

https://github.com/openstreetmap/openstreetmap-website/blob/master/app/controllers/geocoder_controller.rb#L115

Which encodes the bounds as the viewbox parameter which is then passed
through when calling Nominatim.

Tom

--
Tom Hughes (t...@compton.nu)
http://compton.nu/

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

[OSM-talk] Nominatim on the main page

[Paul Johnson]

In the diatribe (no disrespect meant, since it did bring up some relevant
points that we can use constructively, simply a description of the tone)
that emacsen posted last week that we've surely all read by now, he does
bring up one point that I think we can work on for fairly immediate
improvement.

On the OSM homepage, can we use the visible area (or maybe that plus an
exploded offset to a larger surrounding bounds) as a bonding box to be
passed to Nominatim for some context when searching?  This animation really
drives the problem home.

https://gfycat.com/NiceImperturbableCatfish
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk