Re: [SED-ML-discuss] FunctionalRange - terminating conditions

[Andrew M. <ak....@au...>]

On 20/05/13 14:51, Frank T. Bergmann wrote:
> What I have been doing, and how I intended it to be used, was that the functionalRange needed to be parameterized by another range. When I was using it I was usually having this other range be a uniformRange, though of course other ones would work to. In that case it was quite easy to figure out when it would be terminating.
>
> Do you have an example when this would not be feasible?
I think if you have at least one vector or uniform range, that means 
that the number of steps in the simulation experiment are fixed, which 
covers a lot of the experiments people are going to want to do, but it 
is very easy to come up with potential experiments that you could only 
do with non-fixed numbers of steps.

For example, you could describe a fitting which stops when a tolerance 
is reached using a FunctionalRange with a termination condition. 
However, that type of application raises the question of whether we 
really want to make SED-ML task descriptions Turing complete, and 
whether we might be better off adding new tasks / simulations that more 
cleanly describe that type of procedure.

In either case, there really needs to be a bit of text in the 
specification explaining what the restrictions are (e.g. that there must 
be at least one VectorRange or UniformRange) and precisely how SED-ML 
simulators are supposed to process tasks; otherwise, every 
implementation is going to end up handling corner cases differently, and 
we will be back to having non-reproducible simulation experiments.

Best wishes,
Andrew