[OSM-dev] Relationships - working prototype

Frederik Ramm frederik at remote.org
Thu Aug 23 13:47:36 BST 2007


Hi all,

    thank you for discussing & trying the relationship stuff in my  
absence. I was happy to see that many people recognized the potential  
of the concept.

I have noted that discussion has partly focused on turn restrictions,  
with non-relationship ideas proposed by, among others, Andy Robinson  
and Steve Coast. Both these suggestions would work (in a way - I  
don't want to go into too much detail here). But finding some other  
way to describe turn restrictions does not diminish the case for  
relationships in general.

Someone has said that relationships would somehow impose a constraint  
on what people can enter into the system and that this would be un- 
OSM-ish. The opposite is the case; anything that was possible before  
is still possible, and there is much more room for mapping creativity  
- or "unexpected uses"! - with something like relationships in place.

I view the whole thing as an offer to the mapping community - if it  
is not taken up, then it will die and eventually be removed with very  
little harm done. In my eyes, there is very little risk involved with  
just taking relationships on board and seeing how they're used. It is  
not a giant change, really.

I would ask those who oppose the idea of relationships to rethink: If  
you believe that relationships are unnecessary, the best way to prove  
your point is to agree to their deployment and, when the next API  
overhaul is due, just get agreement to drop them based on lack of  
practical use. We do allow people to map all kinds of nonsense using  
fantasy tags, and this is not an accident - we say this over and  
over, OSM is where the mappers decide what to map, not the database  
designers. This is almost an OSM "core value". Relationships will  
furnish the mapper with a powerful new tool to map things that could  
not be properly mapped before, and mappers can use this tool with all  
the degrees of freedom they have applied to tagging in the past. If  
mappers don't use it because other ideas are simpler, easier, faster,  
nicer, or better supported by their editors - fine, they don't have to!


A lot of detailed changes have been suggested and I'll read  
everything and try to make it into a "version 2 prototype" which I'll  
announce here after the weekend. One thing I'm pretty sure of is that  
we need a new name; I suggest we use "relation". ("relationship" or  
"group" would also be ok for me, I don't really mind.) - This word  
does not encompass all possible uses, as the object can also be used  
to represent something by itself, but (a) it'll probably work in 90%  
of cases, and (b) a community that routinely uses something called  
"way" to describe a lake or forest should have no problem with a  
little leeway concerning names ;-)

I would like to enlist your help for the following areas of this sub- 
project:

(a) Is there anyone who is good enough at hacking one of our existig  
renderers - Mapnik, Osmarender, or the Postscript renderer - to  
actually modify one of them to support one or the other kind of  
relationship, a "superway" perhaps? That would make a very cool proof  
of concept.

(b) The same for routing - anyone with a piece of routing software,  
however alpha, who would like to make some modifications to have it  
recognize superways, or turn restrictions or signposting information  
or something?

(c) The idea also needs "political" support. At some point in the not  
so distant future, someone has to actually deploy the stuff on our  
database server, and this is only going to happen if the powers that  
be agree. As most of you know I'm only a very occasional visitor on  
our IRC channel so I cannot argue for relationship support there.  I  
hope that those of you who like the idea will do their part to  
overcome skepticism.

Thanks,
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00.09' E008°23.33'






More information about the dev mailing list