[Geocoding] Namefinder future

[David Earl]

There's a mini discussion going on in one of the trac bug reports for 
namefinder at the moment re "name not found"[1]. It's probably better to 
have the wider discussion here.

The position with Namefinder is unchanged: the index is a year out of 
date. I did try many times to reload the whole planet for a couple of 
months in the summer, but the scale of the task is so much greater than 
when I first did it has meant that it now takes so long as to be 
completely impracticable - weeks, and on the last occasion something in 
the system seems to have restarted after some weeks running.

It is clear that a major amount of work would have to be done to make 
namefinder practical again. I think it would probably work again if the 
index could be reloaded as daily updates aren't particularly any more 
onerous than they used to be. There are a couple of serious but solvable 
problems that need to be dealt with to avoid the updates interfering 
with the searches (but the most obvious that Tom suggested slows the 
whole update process down so much that it wasn't practical. But 
CycleStreets has a pragmatic solution, by swapping database files over 
to make an update available); and searches on more than a few words are 
also problematic, and switching to Popstgres might help there - though 
who knows what other problems that might introduce.

Anyway, there is a lot to do and I am not motivated to do it. If it was 
a matter of tweaking or just sitting there while it reloaded the index, 
fine. But it is apparent that isn't going to work. But I can't see the 
point in me investing the necessary time in Namefinder when Nominatim is 
also doing a good job. No one else has expressed an interest in working 
on Namefinder and the only development response has been to start from 
scratch on a different project. I'd much rather be in a position where 
we are co-operatively working together to produce a better collective 
result than in competition with Brian, and there are so many other 
things I can and am doing that Namefinder is never going to bubble to 
the top if I can't see a good reason for it.

I don't know what the code status of Nominatim is. Is it in svn? Is it 
GPL or similar. There doesn't seem to be a wiki page describing it or 
how it works. Brian, would you welcome other contributions or is it 
something you want to keep to yourself?

There are some things Namefinder does better still: looser matches 
(though it could go much further), contextualising of the result (though 
the OSM home page largely suppresses this); but some things it doesn't 
even attempt (addresses); and it needs to have the syntax for contextual 
searches got rid of. But if it doesn't have a current index, none of 
that has any worth.

I think we should now remove Namefinder from the OSM search results, as 
it's incomplete index is clearly unhelpful and counter-productive. We 
can then free up the machine it runnuing on as well, maybe make it 
available for Nominatim or other things.

Or of someone else wants to get into Namefinder and help improve it, 
that is also a possible way forward.

But as things stand, it isn't sustainable and I'm just not motivated to 
do the necessary work to make it so.

David

[1] http://trac.openstreetmap.org/ticket/1843