[OSM-dev] OSMXAPI to be ported? (Was Re: Lean and mean Tile- and XML-API-Server)

[Frederik Ramm]

Hi,

Stefan Keller wrote:
> So it seems that the OSMXAPI is the candidate which really needs to be ported?

Instead of porting something that exists, why don't you define your 
needs and then simply write something that serves those needs?

OSMXAPI, for example, has a number of special features, e.g. keeping 
track of a list of users having touched an object or "locking" certain 
tags so that only the person having created it can modify it, which are 
cool to have but might be unnecessary baggage for the application you 
have in mind.

It is also important to know whether you're aiming at some sort of live 
replication (using minute diffs and applying them quickly) or whether 
you are aiming at loading daily diffs; this might also cause you to make 
different design decisions. Systems that can work with minute diffs tend 
to spend more time on initial setup, whereas a system that just loads 
the planet file and is thereafter unable to process minute diffs can 
load the planet much faster.

Another factor is whether or not you need to by XML compatible with the 
API; if it is sufficient for your application to hand out, say, WKT 
geometries then you might go down a different route than if you need to 
hand out the exact data objects.

And so on.

If you are aiming at a read/write environment then it will be hard to 
surpass the current API performance. If you intend to do readonly then 
there's a number of shortcuts you can take.

Bye
Frederik

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