[wfs-dev] WFS 1.1, transaction, idgen and its practical applications

[Andrea Aime]

Hi all,
I'm trying to implement wfs 1.1 transaction and I'm having trouble with
the practical applications of idgen in the transaction request.

The spec provides three possible values, GenerateNew, UseExisting, 
ReplaceDuplicate,
and allows the server to specify which values are supported in the 
capabilities document
on a server wide basis.

My problem is that our server (Geoserver) is able to handle a variety of 
data stores
(spatially enabled databases, shapefiles, mif files, gml files and so 
on), and each
data store may have different capabilities to handle different idgen values.
Usually file based data stores are able to use GenerateNew, some work 
only with
UseExisting, and in the case of databases, it really depends on what 
kind of primary
key has been defined in the tables (serial/sequence, single provided 
column, multiple
columns).

So I have the impression I simply cannot implement the spec as is, 
because I would need
to advertise different idgen capabilities for each feature type.

Does anyone has experiences on 1.1 implementation, or are the spec. 
writers willing
to comment somehow this troublesome situation?

Best regards
Andrea Aime