[TinyOWS-users] fids in database views

[Rahkonen Jukka]

Olivier Courtin wrote:


> What about following specs for TinyOWS:
> - Using PK by default if  present (and able to deal with multi columns  
PK)
> - Using incremental fid if no PK founded (maybe using uuid ?)
> - User could overhead the fid by something else than PK (column must  
> be UNIQUE) in config.xml

Sounds good to me.  However, it may be too hard to say "must" in the last case. With views it may be impossible to guarantee that a column is unique.  It can be checked for sure only after running the whole query, which can be very expensive. In Geoserver case the responsibolity is left for the one who configures server. If the defined PK is not really unique, the service generates dublicate fids which can for sure make troubles for WFS clients. I believe it is easy to generate dublicate fids with Mapserver, too. However, dublicate fids are usually not as bad thing for the end user than getting no data at all may be. For example ogr2ogr converts such gml files to totally valid shapefiles if I remember right.

So would it be OK to accept that TinyOWS may generate in some circumstances dublicate fids? If not, should there still be a selection for "Strict" or "Liberal" modes?

-Jukka-