S3 and SimDB

[Gerard]

Thanks Carlos

> In what respect to my use of the word "biased", ...
> ... I once read that 
> someone said: "when the only tool that we have is a hammer, 
> we tend to see all problems as nails" ;-)
> 
Actually the ideas for the model were not influenced originally by
cosmological simulations.
The first "version" of the data model was actually a "domain model" I
developed
together with Pat Dowler, and aimed to give a (not too detailed) description
of astronomy.
It aimed
The ideas of that model were applied to SNAP and seemed to fit.
Others have said they could see it applied also to other types of
simulations.
I was hesitant to try to do so at first because it would remove the focus I
think we need to get SimDB done.
The main idea of the model is the separation of describing *how* you do
experiments (observations, simulations, calculations), for example with a
description of input paarameters, from to the descriptions of the experiment
itself, with parameter settings and results. 
And what I gather so far from S3 is that you (implicitly) use similar ideas.
The protocol in SimDB seems to correspond to the description of your model
in terms of the parameters.
The results of experiments to you "models" which can be queried by the
values assigned to the parameters.you think is 

The SimDB model contains more features to describe the protocol, for example
elements that describe what kind of data products are available. You do that
implicitly in the getData part, namely when defining the fields in the
votable. We believe it may be important to make this available already in
the description of the service.
Not all features in SimDB may be relevant for your type of models, but it
may contain some that 
are not included in your approach and would be useful.

The only aspect in the model that is explicitly targeting the specific type
of simulations
SimDB has focused on (3+1D ones) is in using Snapshot with specific spatial
and time attributes, as results.
This would have to be generalised to also capture models producing other
types of results. It is pretty straightforward to do so.

I agree that the SimDB data model is rather complex, though not necessarily
more complex than
for example the registry model, or STC. And we are trying to find ways
around this to make it
more usable. We (well I and some others) do believe it is a correct
description of the field, 
but it may be more complete than required. One hope was that others could
use it as inspiration for simpler models. Maybe S3 could do something like
that. It would be interesting though to see whether this is true and I'd be
happy to assist in that.


Cheers

Gerard