Excerpts from Tobias Girschick's message of Wed Oct 07 16:00:26 +0200 2009:
> Hello All,
> 
> there is another point that arose during our internal discussions on the
> API. Where does a service that wants to store data, e.g. calculated
> data, know where (or on which server) to store (POST) it? Can we
> introduce something like server_uri as additional parameter for the
> services?
> 
> A possible scenario is that the user calculates features
> using  /algorithm/descriptorcalculator and wants to ensure that the
> generated features (and feature_definitions) are POSTed to a specific
> server, say 
> POST feature123 http://userserver:8080/feature 
> POST feature_definition345 http://userserver:8080/feature_definition
> 
> This issue will/might also arise if a developer does not want to
> implement the whole (or most of) the API, but just a new model
> development or descriptor calculation service as a contribution to
> OpenTox.

Do you think that choosing a dataset server e.g. from a configuration
file is too unflexible for your algorithm service? Your proposal would
mean that each user can select a different dataset server - is this a
common use case? I personally would not offer this choice to my users,
but I think that every developer has the freedom to add parameters to
the existing API (as long as they are documented and the offical API
works with a reasonable default). 

Regards,
Christoph