Re: [SED-ML-discuss] The 'numberOfPoints' attribute

[Lucian S. <luc...@gm...>]

On Wed, Jun 28, 2017 at 2:06 AM, Cooper, Jonathan <j.p...@uc...>
wrote:

> I don’t like the idea of removing an attribute in a version increment – I
> think removing constructs should only happen when the level increases
> really. So I’d like to keep the existing numberOfPoints with its existing
> meaning (albeit perhaps clarified in the spec – but I think the spec was
> already pretty clear on this).
>
> However, I agree that it’s a confusing attribute! So I think I would like
> to see the addition of ‘numberOfSteps’ and users encouraged to use this
> instead.
>

This sounds like my option 3--deprecate 'numberOfPoints' but keep it; add
'numberOfSteps' as a replacement (but disallow both at once).

I do understand the distrust of removing attributes between versions, but
this would still produce versions that were fully compatible with each
other:  all the information in the lower level/versions could be exactly
replicated in the higher level/versions, so we're only adjusting, not
removing any *semantic* information.

The problem with the spec is that it's inconsistent:  the examples and the
text don't always match the description.  It also doesn't help that the
attribute name is a lie, which is the main point of contention.

With regards to the outputTimeInterval, we could have that as well as an
> option if we want. Am I correct in thinking that the current spec allows
> points to be unevenly distributed, thus supporting adaptive integrators
> without needing to interpolate?
>

The problem with adding outputTimeInterval is that it can be used to
overspecify the output, which we presumably don't want.  Any properly
specified output can be translated to min/max/numberofSteps, so hopefully
that could be handled in the user interface of a tool that wanted to use
outputTimeInterval?

-Lucian