Re: [OSM-talk] address parsing by nominatim

2011-01-11 Thread M∡rtin Koppenhoefer
2011/1/11 Stephen Hope slh...@gmail.com:
 2011/1/10 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
 I used hamlet for my block as pop limit of 1000 is given = satisfied

 The problem here is that population is only part of the definition of
 a hamlet.  Less than 1000 people is correct, but it also has an
 implied and is surrounded by open land/farms etc.  You can't have a
 whole bunch of adjacent hamlets sharing borders, they are not hamlets.

+1

the wiki is not very good when defining hamlets (and other places)
IMHO. Although we don't seem to have real problems with that, as
localized versions of the site give different numbers (e.g. hamlet
below 200 in Germany). The main classification for place as it is
handled in the real world (and also the real OSM world) takes into
account different aspects. The number of inhabitants is only a rough
estimate. E.g. a town is a town when it is a town ;-). This is tagged
mainly on a functional and historical (a town in Germany or Italy is a
town by definition, not dependent on the population, I guess this is
the same in most of the world) basis and has to take into account the
surrounding as well. Besides a big city a village can be more then
1 inhabitants, and small towns can have less then 2000 (in
exceptional cases).

A hamlet is different from a village not only because it is smaller,
it also functionally doesn't offer community services (church, shop,
place to gather etc.). The houses are typically not arranged in a
closed way but rather scattered (again, this depends very much on the
cultural/settlement traditions and cannot be generalized, of course
there are also villages with scattered houses). But a hamlet is not a
block or sector of a bigger settlement.

cheers,
Martin

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


Re: [OSM-talk] address parsing by nominatim

2011-01-11 Thread john
A common settlement pattern in the USA, in rural communities that have mostly 
or entirely developed since the invention of the automobile, is to have houses 
and small businesses strung out along a highway.  You end up with a community 
that can be several miles long, and yet only a hundred feet or so wide.  If the 
community was started prior to the automobile, it may have a small portion laid 
out as a grid or loose mesh of streets, with the oldest structures there.  The 
linear portions are where a farmer sold the land immediately adjacent to the 
highway, for house construction, but continued to farm on the rest of his land. 
 The non-farmers are likely to commute to jobs in another, larger community 
nearby.

---Original Email---
Subject :Re: [OSM-talk] address parsing by nominatim
From  :mailto:dieterdre...@gmail.com
Date  :Tue Jan 11 11:31:35 America/Chicago 2011


2011/1/11 Stephen Hope slh...@gmail.com:
 2011/1/10 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
 I used hamlet for my block as pop limit of 1000 is given = satisfied

 The problem here is that population is only part of the definition of
 a hamlet.  Less than 1000 people is correct, but it also has an
 implied and is surrounded by open land/farms etc.  You can't have a
 whole bunch of adjacent hamlets sharing borders, they are not hamlets.

+1

the wiki is not very good when defining hamlets (and other places)
IMHO. Although we don't seem to have real problems with that, as
localized versions of the site give different numbers (e.g. hamlet
below 200 in Germany). The main classification for place as it is
handled in the real world (and also the real OSM world) takes into
account different aspects. The number of inhabitants is only a rough
estimate. E.g. a town is a town when it is a town ;-). This is tagged
mainly on a functional and historical (a town in Germany or Italy is a
town by definition, not dependent on the population, I guess this is
the same in most of the world) basis and has to take into account the
surrounding as well. Besides a big city a village can be more then
1 inhabitants, and small towns can have less then 2000 (in
exceptional cases).

A hamlet is different from a village not only because it is smaller,
it also functionally doesn't offer community services (church, shop,
place to gather etc.). The houses are typically not arranged in a
closed way but rather scattered (again, this depends very much on the
cultural/settlement traditions and cannot be generalized, of course
there are also villages with scattered houses). But a hamlet is not a
block or sector of a bigger settlement.

cheers,
Martin

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

-- 
John F. Eldredge -- j...@jfeldredge.com
Reserve your right to think, for even to think wrongly
is better than not to think at all. -- Hypatia of Alexandria
___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread Vikas Yadav
On 10 January 2011 00:45, M∡rtin Koppenhoefer dieterdre...@gmail.comwrote:

 2011/1/9 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
  I surveying from Northern India. I studied how addresses are to be tagged
 in
  order so nominatim can locate it. That went great. The problem is the
 every
  house has to be tied to a street (addr:street).
  1) We don't have names for living streets


 are you sure you are talking about living streets or do you mean
 residential streets?


  2) We so many times have blocks or sectors (tagged as locality or hamlet)


 locality should be used for uninhabited places

locality is the way i could properly render blocks and sectors right now in
OSM india. please suggest if there is a better/proper way achieving the same
render result.

z12: http://www.openstreetmap.org/?lat=28.408lon=77.0776zoom=12layers=M
 - you can spot sectors (govt sold residential area) as well as private
societies (private sold resi are).

z14: http://www.openstreetmap.org/?lat=28.4141lon=77.0564zoom=14layers=M
now, you can see blocks to sectors too like block A,B,C or Phase 1,2 or Part
1 or Part 2 which are within the same sector. (but there are no known naming
rules so there are always exceptions)
Only sometimes, street names are only connecting sectors. but never are
there street names within a sector.





 sorry, that I cannot help you with your original problem.

 cheers,
 Martin

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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread Gregory Arenius
It sounds like the current set of tags and structures in OSM don't properly
support what you're trying to do at the moment.  I think the best way to
deal with it would to propose some new tags so that these type of addressing
schemes can be properly supported in OSM with supporting documentation.

It sounds like you need some sort of block or sector tag you could set on
areas to define them properly.  A couple of other tags for addressing like
addr:block and addr:sector could then be added.

It would definitely take some time before they get supported by the
renderers and routers but I think that since a good sized chunk of the world
uses these types of addressing systems we should have an explicit way to
deal with them instead of trying to shoehorn them into a more European
system.  It will make it easier for mappers there to do things properly and
probably also make things work more reliably.

I don't have the experience with this type addressing to feel comfortable
doing any of that but if you put something up on the wiki people can use and
work on it.

Cheers,
Greg
___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread vikas yadav
Neither block or sector are OSM places. I have used halmet/locality and
suburb. by these, rendering is proper
if using the existing system could i propose addr:hamlet or addr:suburb
support?
Also, rendering probably is fine.
Its just how gazetteer/nominatim search algo that would require supporting
alternate tags.

Thanks,
Vikas

On 10 January 2011 15:34, Gregory Arenius greg...@arenius.com wrote:

 It sounds like the current set of tags and structures in OSM don't properly
 support what you're trying to do at the moment.  I think the best way to
 deal with it would to propose some new tags so that these type of addressing
 schemes can be properly supported in OSM with supporting documentation.

 It sounds like you need some sort of block or sector tag you could set on
 areas to define them properly.  A couple of other tags for addressing like
 addr:block and addr:sector could then be added.

 It would definitely take some time before they get supported by the
 renderers and routers but I think that since a good sized chunk of the world
 uses these types of addressing systems we should have an explicit way to
 deal with them instead of trying to shoehorn them into a more European
 system.  It will make it easier for mappers there to do things properly and
 probably also make things work more reliably.

 I don't have the experience with this type addressing to feel comfortable
 doing any of that but if you put something up on the wiki people can use and
 work on it.

 Cheers,
 Greg



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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread Jacek Konieczny
On Mon, Jan 10, 2011 at 04:04:51PM +0530, ヴィカス ヤダヴァ (vikas yadav) wrote:
Neither block or sector are OSM places. I have used halmet/locality and
suburb. by these, rendering is proper
if using the existing system could i propose addr:hamlet or addr:suburb
support?

And why not is_in:*, including recursive is_in:* (something has
is_in:suburb=something, something suburb has is_in:city=something_else,
etc), probably used together with addr:*? I know, that administrative
boundaries would often be better, but those are often not available.

Greets,
Jacek

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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread M∡rtin Koppenhoefer
2011/1/10 Vikas Yadav vi...@thevikas.com:
 On 10 January 2011 00:45, M∡rtin Koppenhoefer dieterdre...@gmail.com
 wrote:

 2011/1/9 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
  I surveying from Northern India. I studied how addresses are to be
  tagged in
  order so nominatim can locate it. That went great. The problem is the
  every
  house has to be tied to a street (addr:street).
  2) We so many times have blocks or sectors (tagged as locality or
  hamlet)
 locality should be used for uninhabited places

 locality is the way i could properly render blocks and sectors right now in
 OSM india. please suggest if there is a better/proper way achieving the same
 render result.


This is what we call tagging for the renderers = abusing a tag in a
way it is not describing what it is defined for, just because this
renders a nice picture. If you need another place-tag, that is not
documented here:

http://wiki.openstreetmap.org/wiki/Place

(e.g. place=block / sector or whatever), you invent it. You can use
all tags you like, but it is recommended for features of which you
think that they are valuable for other users as well, to document them
in the wiki (and maybe send a note to the mailing lists). Usually you
would make a proposal.

cheers,
Martin

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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread Brian Quinion
 2) We so many times have blocks or sectors (tagged as locality or hamlet)

OK - this sounds like a combination of bad tagging and software
problems with nominatim and the mapnik style sheet.

Can I suggest that locality and hamlet are probably not the correct
tags and that you need to come up with a consistent way to tag these
types of features.  Once there is a valid tagging scheme support can
be added to nominatim.

The suggestion of place=block or place=sector and addr:block,
addr:sector sound like a good direction to go.

So

1) they need to be documented on the wiki
2) discussed to look for any problems - for instance do they work in
other countries with similar problems?
3) a sample area needs to be tagged
4) software tools need to be extended to support the new tags.

This can potentially be done quite quickly - and using tags which do
not overlap with existing tags makes this a LOT simpler!

--
 Brian

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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread vikas yadav
I used hamlet for my block as pop limit of 1000 is given = satisfied
I used suburb for it is neither a village or a town but holds 2 ~ 10 blocks
= suggest

We do have villages within cities and they have been tagged properly,
villages never have sub areas or blocks.
Therefore, sectors are not villages.

On 10 January 2011 16:35, Brian Quinion
openstreet...@brian.quinion.co.ukwrote:

  2) We so many times have blocks or sectors (tagged as locality or hamlet)

 OK - this sounds like a combination of bad tagging and software
 problems with nominatim and the mapnik style sheet.

 Can I suggest that locality and hamlet are probably not the correct
 tags and that you need to come up with a consistent way to tag these
 types of features.  Once there is a valid tagging scheme support can
 be added to nominatim.

 The suggestion of place=block or place=sector and addr:block,
 addr:sector sound like a good direction to go.

 So

 1) they need to be documented on the wiki
 2) discussed to look for any problems - for instance do they work in
 other countries with similar problems?
 3) a sample area needs to be tagged
 4) software tools need to be extended to support the new tags.

 This can potentially be done quite quickly - and using tags which do
 not overlap with existing tags makes this a LOT simpler!

 --
  Brian

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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread Pierre-Alain Dorange
Vikas Yadav vi...@thevikas.com wrote:

 locality is the way i could properly render blocks and sectors right now in
 OSM india. please suggest if there is a better/proper way achieving the same
 render result.
 
 z12: http://www.openstreetmap.org/?lat=28.408lon=77.0776zoom=12layers=M
  - you can spot sectors (govt sold residential area) as well as private
 societies (private sold resi are).
 
 z14: http://www.openstreetmap.org/?lat=28.4141lon=77.0564zoom=14layers=M
 now, you can see blocks to sectors too like block A,B,C or Phase 1,2 or Part
 1 or Part 2 which are within the same sector. (but there are no known naming
 rules so there are always exceptions)
 Only sometimes, street names are only connecting sectors. but never are
 there street names within a sector.

I've nor followed the whole discussion but perhaps can you test using
boundary relation. You can define area with name and nominatim use them.

relation
type=boundary
boundary=administrative
admin_level= ?

http://wiki.openstreetmap.org/wiki/Admin_level
it seems 6 for blocks

but the situation in India seems a bit tricky :
http://en.wikipedia.org/wiki/Administrative_divisions_of_India



-- 
Pierre-Alain Dorange
OSM experiences : http://www.leretourdelautruche.com/map/


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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread Brian Quinion
2011/1/10 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
 I used hamlet for my block as pop limit of 1000 is given = satisfied
 I used suburb for it is neither a village or a town but holds 2 ~ 10 blocks
 = suggest

 We do have villages within cities and they have been tagged properly,
 villages never have sub areas or blocks.
 Therefore, sectors are not villages.

You seem  to be determined to force the existing tagging scheme onto a
situation for which is was not designed.  It as far better to use new
and appropriate tags that reflect the actual situation - software
support should follow fairly rapidly if you come up with a suitable
tagging scheme.

Using hamlets, villages and incorrectly named roads to try to hack the
various software will not work and is very unlikely to be supported by
any of the software.

Pierre-Alain Dorange's suggestion to use admin_level and some type of
boundary is another reasonable way to approach the problem - although
I would suggest that your probably want to create something like
admin_level=11 or maybe even 12.  6 Is definitely too low.

You may also be able to find inspiration in the tagging of Japan - my
understanding is that they have a similar approach and may already
have created a suitable scheme.

--
 Brian

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


Re: [OSM-talk] address parsing by nominatim

2011-01-10 Thread Stephen Hope
2011/1/10 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
 I used hamlet for my block as pop limit of 1000 is given = satisfied

The problem here is that population is only part of the definition of
a hamlet.  Less than 1000 people is correct, but it also has an
implied and is surrounded by open land/farms etc.  You can't have a
whole bunch of adjacent hamlets sharing borders, they are not hamlets.

Stephen

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


[OSM-talk] address parsing by nominatim

2011-01-09 Thread vikas yadav
Hi,

I surveying from Northern India. I studied how addresses are to be tagged in
order so nominatim can locate it. That went great. The problem is the every
house has to be tied to a street (addr:street).
1) We don't have names for living streets
2) We so many times have blocks or sectors (tagged as locality or hamlet)

Just to get the parser work accurately I had to name closeby living_street
street as Sector 46 or Block B - which when rendered would look very
strange cause that should only be name of the area and not the name of
street. But that is the only way I could successfully parse address such as
19, Block A1, South City II, Gurgoan or 1532, Sector 46, Gurgoan which
are exact postal address.

So is there a way that I can tie a addr: to a hamlet/locality instead of
street?

(I installed Nominatim myself to be able to hit-and-try and learn the exact
way, therefore these example addresses won't work on live on current data.)

Please suggest,
Vikas
___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] address parsing by nominatim

2011-01-09 Thread M∡rtin Koppenhoefer
2011/1/9 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
 I surveying from Northern India. I studied how addresses are to be tagged in
 order so nominatim can locate it. That went great. The problem is the every
 house has to be tied to a street (addr:street).
 1) We don't have names for living streets


are you sure you are talking about living streets or do you mean
residential streets?


 2) We so many times have blocks or sectors (tagged as locality or hamlet)


locality should be used for uninhabited places


sorry, that I cannot help you with your original problem.

cheers,
Martin

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


Re: [OSM-talk] address parsing by nominatim

2011-01-09 Thread Elizabeth Dodd
On Sun, 9 Jan 2011 20:15:46 +0100
M∡rtin Koppenhoefer dieterdre...@gmail.com wrote:

 2011/1/9 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
  I surveying from Northern India. I studied how addresses are to be
  tagged in order so nominatim can locate it. That went great. The
  problem is the every house has to be tied to a street (addr:street).
  1) We don't have names for living streets
 
 
 are you sure you are talking about living streets or do you mean
 residential streets?
 
 
  2) We so many times have blocks or sectors (tagged as locality or
  hamlet)
 
 
 locality should be used for uninhabited places
 
 
 sorry, that I cannot help you with your original problem.
 
 cheers,
 Martin
 

Martin's reply highlights a problem with the eurocentric version of a
place hierarchy in OSM - it doesn't fit India.

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


Re: [OSM-talk] address parsing by nominatim

2011-01-09 Thread Daniel Kastl
2011/1/10 Elizabeth Dodd ed...@billiau.net

 On Sun, 9 Jan 2011 20:15:46 +0100
 M∡rtin Koppenhoefer dieterdre...@gmail.com wrote:

  2011/1/9 ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com:
   I surveying from Northern India. I studied how addresses are to be
   tagged in order so nominatim can locate it. That went great. The
   problem is the every house has to be tied to a street (addr:street).
   1) We don't have names for living streets
 
 
  are you sure you are talking about living streets or do you mean
  residential streets?
 
 
   2) We so many times have blocks or sectors (tagged as locality or
   hamlet)
 
 
  locality should be used for uninhabited places
 
 
  sorry, that I cannot help you with your original problem.
 
  cheers,
  Martin
 

 Martin's reply highlights a problem with the eurocentric version of a
 place hierarchy in OSM - it doesn't fit India.


Not only India, also other countries such as Japan have a block address
system, which for example doesn't match with how openaddresses collects data
at the moment.
Here some good explanation: http://sivers.org/jadr

Such a block address system actually has some advantages as well. You don't
have to know the house number to be able to know about where the address is
located. You can take the center point of the block polygon as well and it
will usually be a good approach. With long streets this is usually not
possible.

Daniel





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




-- 
Georepublic UG  Georepublic Japan
eMail: daniel.ka...@georepublic.de
Web: http://georepublic.de
___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


[OSM-ja] Fwd: [OSM-talk] address parsing by nominatim

2011-01-09 Thread Hiroshi Miura
三浦です。

インドのVikasさんが、こんな投稿をしています。
日本と共通の課題ではないですか?


 Original Message 
Subject:[OSM-talk] address parsing by nominatim
Date:   Sun, 9 Jan 2011 17:44:46 +0530
From:   ヴィカス ヤダヴァ (vikas yadav) mevi...@gmail.com
Reply-To:   vi...@thevikas.com
To: osm-talk t...@openstreetmap.org



Hi,

I surveying from Northern India. I studied how addresses are to be
tagged in order so nominatim can locate it. That went great. The problem
is the every house has to be tied to a street (addr:street).
1) We don't have names for living streets
2) We so many times have blocks or sectors (tagged as locality or hamlet)

Just to get the parser work accurately I had to name closeby
living_street street as Sector 46 or Block B - which when rendered
would look very strange cause that should only be name of the area and
not the name of street. But that is the only way I could successfully
parse address such as 19, Block A1, South City II, Gurgoan or 1532,
Sector 46, Gurgoan which are exact postal address.

So is there a way that I can tie a addr: to a hamlet/locality instead of
street?

(I installed Nominatim myself to be able to hit-and-try and learn the
exact way, therefore these example addresses won't work on live on
current data.)

Please suggest,
Vikas

--
Ovi Mail: Making email access easy
http://mail.ovi.com


___
Talk-ja mailing list
Talk-ja@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-ja