[kepler-dev] Kepler Alpha8

[Bertram Ludaescher]

>>> On Mon, 19 Dec 2005 12:44:14 -0800
>>> Dan Higgins <higgins at nceas.ucsb.edu> wrote: 
DH> 
DH> Hi All,
DH>     There is a zipped version of the potential alpha8 Kepler release at
DH> http://www.kepler-project.org/dist
DH> under the name 'kepler-alpha8_test1.zip'
DH> 
DH> You may well want to take a look. Windows users should be able to run 
DH> the code using 'kepler.bat' (and Linux/Mac users using kepler.sh). Note 
DH> that the zipped version does not use or require preset environment 
DH> variables, so you should be able to unzip anywhere. Also note that the 
DH> PTII directory is not needed since compile PTII classes are in the 
DH> kepler.jar file.
DH> 
DH> We have several unresolved problems. First, any of the 'IPCC' data 
DH> sources cannot be dragged onto the work area (SRB problem?)
DH> 
DH> Is the PIW workflow working?
DH> 
DH> Are the GEON workflows working? (web services - are they working?)
DH> 
DH> It looks to me like the ORB example workflows do not work.
DH> 
DH> Any comments?

Two comments:

(1) Can we assign an "owner" to each demo-workflow? Then the owner
would be in charge of making sure everything runs, and if it doesn't
the workflow could be move from the "demo collection" (part of the
package) to the "experimental collection" (probably not part of the
package, but dynamically installable.. via "the Kepler repository")

(2) Inherently, complex workflows using many non-local resources (web
services, data sets etc) are more fragile than simple ones (or complex
ones with few non-local resources). Thus, we might want to think about
refactoring some of the complex ones into smaller pieces -- for demo
purposes. Demos should demonstrate features, not be complex
"production workflows".

(3) Not sure whether caching could be used for demos in Kepler. I've
been involved in two other systems (Florid and the early BIRN
mediator) whihc had a result caching mechanism. So you could prepare a
demo while online, then show it offline, i.e., running from the
cache. If the current caching system allows such a use (or misuse some
might say ;-) then this is another way to make demos more resilient.. 
This mechanism could also double for automatically checking
(deterministic!) workflows in nightly test suites..

Bertram

DH> 
DH> Dan
DH> 
DH> 
DH> 
DH> 
DH> -- 
DH> *******************************************************************
DH> Dan Higgins                                  higgins at nceas.ucsb.edu
DH> http://www.nceas.ucsb.edu/    Ph: 805-893-5127
DH> National Center for Ecological Analysis and Synthesis (NCEAS) Marine Science Building - Room 3405
DH> Santa Barbara, CA 93195
DH> *******************************************************************
DH> 
DH> 
DH> _______________________________________________
DH> Kepler-dev mailing list
DH> Kepler-dev at ecoinformatics.org
DH> http://mercury.nceas.ucsb.edu/ecoinformatics/mailman/listinfo/kepler-dev