[OSM-dev] 0.6 api - one more time

[Richard Fairhurst]

SteveC wrote:

> I disagree, publicise widely and just switch. It's worked every time
> in the past. It's not _entirely_ clear that it's possible to run both,
> and in the past it certainly hasn't been such as the dropping of
> segments.
>
> The only think I have a problem with is backward supporting the crap
> in potlatch.

No-one has ever needed to add "backward supporting" code for Potlatch  
anywhere, whether on 0.4 or 0.5, and there's no question of them doing  
so this time, either. You may not like the internals and I can  
understand that, but Potlatch always tries to play nice. With 0.6, as  
with 0.5, the main API is actually moving slightly closer to the way  
Potlatch has always done things _anyway_.

Ah, the bullets you have to take for being ahead of the curve. :p

But to be a bit more helpful:

> is there anything else that should be put in 0.6? Probably a good  
> idea to talk about it now.

Well, ideally it'd be good to add

PUT /api/0.6/way/123456/full
to do the same thing as Potlatch's 'putway' call, and then migrate the  
latter to use the same methods.

It would also be good to make
way.delete_with_relations_and_nodes_and_history

actually cope with relations - at present it just throws a  
Precondition Failed if you try to delete a way which is a member of  
any relations.

Those two are the bulk of the SQL in amf_controller. Fixing those  
would just leave the POI stuff (which is sufficiently trivial I can  
probably have a go at it without breaking too much) and Potlatch's  
atomic history/revert, where I suspect the best strategy is to leave  
it as is until we've seen how people are using 0.6-style revert in  
practice.

cheers
Richard