CCA Fall Meeting Minutes

A serious amount of stuff got done. I think a serious amount of stuff had to get done*1.

Here are the results:

  1. Steve lead us through some simple (and hopefully final) changes to the core specification. This gets rid of createPortInfo in favor of putting the information as arguments to addProvidesPort() and registerUsesPort(). Steve (I believe) will post this to the voting machine at CASC.
  2. It appears that the work is done on BuilderServices Port and AbstractFramework. These interfaces make it possible to build applications in CCA and mark a deviation from a strict "components only" model. Ben, Dennis, Steve and myself all contributed to this. There were some misgivings from some quarters (Dennis and especially Steve) regarding the potential for abuse of these new interfaces. I will volunteer to put the results of our discussions on the voting machine.
  3. Gary and Scott gave the view from SIDL and Gary gave an implementation of a CCA framework in SIDL and emphasized the need for and 2 above. Some discussion was had on the difficulty of packaging components and whether or not SIDL could provide that service.
  4. Jeeem representing the MxN subgroup, recognizing that the MxN interfaces presented meeting-before-last were far too complex, and recognizing that the simplified interfaces presented at the next meeting were far too simple, presented a new set that were both simpler and more complex ;-). This was the result of lengthy email discussions with the MxN subgroup. They plan to have a face-to-face meeting the day prior to the next CCA meeting (see below).
  5. Craig talked a little about how to deal with F90 in a mixed language environment and Matt presented mechanisms for wrapping C++ in SIDL. This is of significant interest to the people doing C++ components and frameworks.
  6. Kate Keahey brought up the idea of a tutorial for those new to CCA before the next CCA regular meeting. This idea was well received and Kate will coordinate.

Data Component SubGroup

The Data Component Subgroup met the afternoon before the CCA regular meeting. The Data Component Subgroup files this report:

At the Scientific Data Component Meeting we did the following:

I presented the first round of the TSTT interfaces for querying unstructured meshes and received feedback from CCA participants regarding improvements and additional functionalities to consider supporting. David Bernholdt presented an overview of the local array and distributed data interfaces. The group then split into several subroups to discuss the details of the various interfaces.

I propose to take a brief hiatus from the data subgroup meetings so that other time critical groups - in particular the MxN subgroup (which overlaps significantly with the data subgroup) and the CCA tutorial development group (also significantly overlapping) can meet without extending the length of the CCA meeting to a full week.

Next meeting: Santa Fe, NM, 10-11 Jan '02

The MxN meeting is scheduled for the morning of 9 Jan '02, probably at the same place or nearby in Santa Fe. We have tentatively scheduled a CCA tutorial for newbies for that afternoon.


*1Note: this text has been formatted in RD <URL:http://www.rubycentral.com/book/rdtool.html> to produce decent web output see: <URL:http://www.cca-forum.org/mtg/2Oct01/minutes.html> while still being generally readable.