[kepler-dev] Re: [SDM-SPA] SPA configuration

Edward A Lee eal at eecs.berkeley.edu
Thu Apr 1 19:41:12 PST 2004

This seems like exactly the right approach, but with one
addition that I would suggest:

Make it easy to publish actor libraries.

A lot of software systems fail because they provide infinite
configurability, but unusable defaults (the X window system,
for example).  This works fine for someone who really wants
to spend several hours configuring the system to their tastes,
but it doesn't work for most people.  I think that if, say,
someone in the oceanography community were to develop a useful
library, then we should make it very easy to someone else to
use it...


At 05:37 PM 4/1/2004 -0800, Bertram Ludaescher wrote:

>Yes, it is good idea to have a "simpler configuration" for actor and
>director libraries.
>However, I don't think that Ilkay (or anyone else from us "IT folks")
>should or even can determine which actors/directors "must" remain and
>which ones can go.
>First, SDM Center/SPA was reviewed a year ago or so, and the critique
>was that we look like "consultants" for a specific community (actually
>specific guy = MattC).
>So actors who are not needed today, may be useful tomorrow. The
>*scientists* need to be able to select and view actor and director
>libraries in a convenient way.
>As I explained in an earlier email to SDM-SPA
>one way would be to annotate actors (and data sets for that matter)
>with a list of properties and the dynamically create a browsable
>folder tree which is organized according to the properties that the
>scientist user has selected.
>This will be much more productive and innovative than trying to second
>guess what is redundant (today) and may be needed later. For example,
>in addition to catering to MattC's needs (bioinformatics), NCSU is now
>interested in suppporting an astrophysics workflow. Maybe actors that
>look useless to us today would be useful for that one tomorrow.
>Other domains for which we already have workflows (sometimes with SPA
>support) are ecology (GARP species predication pipeline), geosciences
>(mineral classification workflows), and cheminformatics (sth coming
>So I think having a dynamic (actor and director) library viewer that
>can be configured by the user is a better solution to the "simpler
>configuration problem" that you mention.
>Note that in Kepler such a mechanism is already on the to do
>list. Specifically, the actor/directory libraries will be dynamically
>fed from a remote repository (e.g., a web service repository), so the
>whole library thing will get much more dynamic than the current static
>folder structure.
>To get an idea how this may look, you may check out the Cmap tools:
>         http://cmap.ihmc.us/download/
>once you install that, you can access "on demand folders" from the web
>(which in this case, have concept maps, but in our case it would be
>actors and data sets...)
>Note that there are also some design documents in the Kepler CVS
>describing an initial design:
>(have a look at the stuff by Jing *Tao* here:
>and here:
>Bottom line: instead of worrying about what actors to delete and how
>to restructure directories, I'd like to suggest to add a more
>sophisticated capability that lets the user him/herself configure the
>view they get. It's already in the making...
> >>>>> "XX" == Xiaowen Xin <xin2 at llnl.gov> writes:
>XX> Hi Ilkay,
>XX> It may be useful to have SPA start up with a simpler configuration than
>XX> the current default.  For example, there are now 8 directors listed in
>XX> the treeview, and our workflows only use one or two of them.  Also, I
>XX> don't believe SPA uses any of the libraries under "more libraries", many
>XX> of them are optional packages that aren't even installed.
>XX> If you start ptolemy with the -ptiny flag, then it uses a smaller
>XX> configuration,
>XX> The abundance of actors that aren't directly related to SPA may cause
>XX> confusion for our users, so we may want to create our own SPA
>XX> configuration.  Which actors do you believe _must_ remain there, and
>XX> which do you think we can remove from the treeview?
>XX> Thanks,
>XX> Xiaowen
>kepler-dev mailing list
>kepler-dev at ecoinformatics.org

Edward A. Lee, Professor
518 Cory Hall, UC Berkeley, Berkeley, CA 94720
phone: 510-642-0455, fax: 510-642-2739
eal at eecs.Berkeley.EDU, http://ptolemy.eecs.berkeley.edu/~eal

More information about the Kepler-dev mailing list