another iMesh / fields issue we may want to consider.

[Mark Shephard]

Sounds right to me.

Carl Ollivier-Gooch wrote:
> On 10-05-07 06:24 AM, Mark Shephard wrote:
>> At the boot camp there was extensive discussion on the issue of where
>> mesh shape information and coordinate information goes. There was
>> general consensus that if we were starting at square one all of this
>> would be in fields. However, because where iMesh is already, that what
>> "is in iMesh" will remain there, but that fields will be able to fully
>> support all mesh shape and coordinate information.
> 
> See below.
> 
>> The item I wanted to mention here is that there was a couple of side
>> comments on the "higher order node stuff" including, if I recall
>> correctly, Tim asking if what we had defined could be revisited. With
>> all the other stuff we were discussing at the boot camp we did not go
>> any deeper into that and I was wondering if we should. I know that Carl
>> and Luo had worked pretty hard on getting something into iMesh for
>> higher order nodes that attempted to address the concerns of Tim and
>> SCOREC. However, I am not sure how much of that was implemented and/or
>> how happy people were with the compromise.
> 
> I'm pretty sure no one ever implemented any of it, unless it was someone 
> at SCOREC.  I know I didn't, and I don't think Tim did, either.
> 
>> If the above is the situation with respect to this issue, the question I
>> would ask is as follows: Should we consider having iMesh support just
>> the coordinate information originally defined and have what we do in the
>> fields stuff deal with all the additional possibilities?
> 
> One of the pieces of homework I have from the bootcamp is to look at the 
> pros and cons of keeping linear mesh shape in iMesh versus moving 
> everything to iField.  Once we have that analysis in front of us, we can 
> decide what to do about that.
> 
> I don't see any future in putting the curved mesh shape stuff into 
> iMesh, considering it would need to provide much the same functionality 
> as iField does.  There is almost certainly some stuff in that API that 
> we can mine (especially the p-version stuff), I just haven't had a 
> chance to look at what things can be transferred / generalized.
> 
> So in the end, I think the answer will almost certainly turn out to be 
> that fancy coordinate stuff will all happen in iField.  Whether simple 
> coordinate stuff will also be supported in iMesh is an open question.
> 
> Carl
>