Talk:Variable transfer protocol

Discussion about model connectivity

Dear Jouni,

I believe the key word here is SOA, service oriented architectures. It basically means connections over protocols like http, and request encoding and data exchange using XML. The key thing is to agree on the data structures and their semantics, meaning how you name things, in the XML. In many domains XML schema (the structure + naming) are agreed upon by standards bodies, such as ISO.

I'm most familiar with geographical data. If you look at the OGC site (http://www.opengeospatial.org/ ) you'll see a large number of (ISO, established, proposed) standards for representing spatial data, exchanging them, accessing them etc. New developments are e.g. services for sensor observations (e.g. to request one or more air quality measurements), or web processing services that can run any (spatial) model on request, on a remote machine. A possibility here is that the request involves uploading (or pointing to) the input data, or using available input data on the machine running the service.

In the context of the INTAMAP project we develop a web service for automatic interpolation, and we are still in search for a good standard (schema) for exchanging a simple thing like a probability distribution function. They might be somewhere in a completely different domain, such as biology; there is a MathML, but it doesn't do it well. We may need to write one, it is not so hard. Making software components work together over a distributed network is usually much harder.

I hope this make sense. If you need any further references, let me know.

Best wishes, -- Edzer (11.9.2007)

Jouni Tuomisto wrote:

Dear all,

Greetings from a Heimtsa project meeting from last week. Our sister project is well speeding up, and the partners there (quite a lot of overlap with Intarese) are actively planning policy scenario modelling. There was one thing there that really was thought-provoking to me. I'd like to share these thoughts and seek your help, as many of you have much more modelling expertise than I do.

Heimtsa is more practically oriented than Intarese (which is intended). Because of this, there were very interesting discussions about how to do complex impact pathway (or full-chain) modelling in practice in a situation where different parts of a large model are computed by different partners. Denis had this very nice idea about systemtising this work. Basically, it was an idea that different institutes would have, develop, and run their own modules that belong as parts to a large impact pathway model. The inputs needed for a module, and the outputs from the module, should somehow be made available to other parts of the whole model system.

The big question is, how this can be done. There should be some standard interfaces for connecting modules, or in other words, standards to make sure that one module has an output that is an appropriate input for another module. The Intarese toolbox gives little help to this at its current stage. However, I believe that there is soon a clear need to make progress in this area in Heimtsa. Whether the development is done in Heimtsa or Intarese, I don't know. But it would be very good if these functionalities could be incorporated into the Intarese toolbox in the future.

Does anyone have good ideas on how to do this combining work with modules? I'd be really interested in participating in this discussion.

Sincerely,

Jouni