|
Jan 16, 2008
Participants: Altintas, Barseghian, Crawl, Cornillon, Gallagher, Graham, Hosseini, Jones, Potter, Schildhauer
- REAP Meeting
- Current idea: UCSD, March 24-27, 2008 (travel 24th, leave afternoon or eve of 27th)
- Tentative plan:
- 1st half) devs work with domain scientists to further implementations of workflows and gui designs
- 2nd half) devs break off and code, while domain scientists continue the above with e.g. Matt and/or Mark
- Goals:
- discuss and progress on datasource actors, rbnb link
- progress on usecase workflows
- brainstorm session on Engineering View (monitor and control of sensors and sensor networks)
- Jones
- bear in mind we have budget for smaller ad-hoc meetings too
- marratech teleconferencing software an option we've been trialing (video, whiteboard) (link to come)
- OPeNDAP discovery and subsetting
 discussion
- Potter explains his work so far. Should I proceed?
- Discussion. Consensus: yes.
- Hosseini: Nodes of different types differentiated in some way, instructions apparent in one of the panes.
- Cornillion: I'd like geospatial and temporal searching; try to reuse existing search motifs.
- Cornillion, Potter to iterate on this interface, and then send to group for review. Get an OPeNDAP datasource actor working in this manner, a second pass for consolidation with other datasource actors can then be done.
- Jones: How to subset and use large number of datasources?
- Cornillion: check how ODC let's you subset, it does it nicely.
- Ocean workflow
- Schildhauer: should we also try to implement Huai-Min Zhang's code and Podesta's code in Kepler?
- Cornillion: this would be a great start.
- Matlab actor discussion
- Altintas/Crawl: Matlab issues not resolved yet-- doesn't take output and serialize it back into Kepler.
- not resolved, but is a high on todo list. Will likely try to use Ptolemy version of this actor.
- Jones: keeping eg Matlab session open across a workflow speeds things up but violates workflow system assumptions (actor independence). Must be aware of this if we make it an option (like the R actor), can run into problems with e.g. distributed computing and PN director which uses threads.
- Altintas: hopefully finish this in a month.
- Todo
- Altintas and Barseghian - work with domain scientists to progress on initial workflows before San Diego
- Crawl - enter and maintain a bug on Matlab actor
- Potter and Cornillion - continue work on OPeNDAP UI, send to group for review
|
