[kepler-dev] "getting started" questions

[Kenneth Jones]

Chad,

To tell you the truth, I didn't check very closely for error messages when I
did the build.  Everything compiled, so I thought I was golden.  It is very
likely that this was nothing more than a case of pilot error.  I'll do the
build again and let you know if I have any difficulties.

I want to try to build a system I envisioned the late 80's/early 90's.  At
the time I was working for a company called Thinking Machines that did some
pioneering work in massively parallel computing.  I was working on the use
of massively parallel computing in scientific and engineering applications
(I have a Ph.D. in physics and a strong background in computing).  I was
working with users who were interesting in solving problems in their domains
and to them computing was nothing more than a necessary evil.  I wanted to
create a system that was as simple as possible for them to use.  I had some
concrete ideas about how to go about creating the system, but the building
blocks I needed simply weren't available at the time and I realized that
what I wanted to do was not something that a single person could create.  So
I waited patiently and kept tabs on the evolution of the technologies I
would need to build the system and I think the time to try to knit together
the pieces has finally arrived.  It seems to me that the Kepler team has
some of the same goals I have.  I am now doing business-oriented computing
and my goals have shifted to the creation of a system that business people
find easy to use, but the basic ideas are essentially the same; it's just
the domain that has changed.

I want to try to build a system that allows a nonprogrammer to create an
application by using a tightly integrated mix of a controlled natural
language and a visual language.  The visuall language will employs actors
and data types that are created by programmers (initially I plan to focus on
the creation of workflows with SDF semantics, but other domains of
computaion might be added later).  The glue that I want to use to bind the
two representations together is logic, imperative commands with functional
semantics, and queries that are expressed in full classical FOL when
neccesary and, when they offer sufficient expressivity, the more restrictive
description logics.  I would like to use a customized version of Kepler for
the framework that supports the definition of actors and data types and the
syntactical and semantics aspects of the visual language.  I plan to use the
work of the Attempto group
<http://attempto.ifi.uzh.ch/site/description/>for the framework that
supports the definition of the lexical elements
(which correspond to Kepler's tokens and actors), syntax definition,
parsing, and translation to Discourse Representation Structures (which can
be translated to statements in FOL) and OWL of a controlled natural
language.  If, in my attempt to create this system, I am able to create
anything useful to either the Attempto group are the Kepler project, I would
like to donate it to the appropriate project.

With these goals in mind, I was very happy to see that the Kepler team had
moved from Ptolemy II's XML format for actor classification to one based on
OWL.  I would like to see if I can develop my software to conform to your
achitectural plans and coding style so I can donate anything you might find
useful.  I have found the style guide on the Kepler website, but I get the
impression that some elements of the architecture are in a state of flux.
Might I be able to make a contribution?  Are there some docs that outline
your goals and what you've done so far?  Are there specific classes I should
study?  Can you point me to the prototyping you've done, even if the code
doesn't work?  Are there other specific team members I should correspond
with?  If there's anything I can do to make a contribution, please let me
know.  I also understand that the timing might not be right because the team
might have it's hands full preparing for the 2.0 release.  If that's the
case, let me know and I'll wait to bug you until after the release.

Thanks,

-- Ken

On Wed, Jun 3, 2009 at 12:42 PM, Chad Berkley <berkley at nceas.ucsb.edu>wrote:

> Hi Ken,
>
> Were you running off of a kepler-1.0 release checked out from SVN or one
> you downloaded (with and installer) from the web?  In any case, it's weird
> that you didn't have those files unless maybe the checkout failed half way
> through or something.
>
> We're not currently doing any reasoning with the actor library ontologies,
> but we're laying the groundwork for kepler to be able to suggest possible
> workflow component connections between actors and to do semantic type
> checking between inputs and outputs.  Some of this functionality has been
> implemented as prototypes, though I'm not sure how much (if any) of it
> currently works.
>
> The plan originally was to let different domains classify their actors the
> way that made sense to them in the actor library.  Since we haven't had the
> tools to easily create the ontologies required, it's been hard to get people
> to actually build new classifications.  The 4 ontologies currently in kepler
> will stay and possibly more will be added.  I could also see in the future
> that the ontologies themselves could be uploaded to the component repository
> and shared amongst users.  I don't think there are any concrete plans for
> that though.
>
> We were trying to shoot for the middle of June for the 2.0 release, but
> it's probably not going to happen as we envisioned it.  I'm guessing the end
> of the summer or fall at this point, however, we should have some alpha and
> beta releases sooner.
>
> Let me know if you have any other questions.
>
> chad
>
>
> Kenneth Jones wrote:
>
>> Chad,
>>
>> Thanks for the timely and detailed response.
>>
>> None of the actors display icons and the reason seems to be that I don't
>> have a svgIconDefaults.properties file in any of the projects in my
>> workspace.  I see that you have created a newer, more detailed, set of build
>> instructions and pointed to them in the most recent Kepler Newsletter.  I
>> think I'll wipe out my current workspace and start again from scratch using
>> the most current build instructions.  That may do the trick.
>>
>> The fact that you are using OWL to categorize your actors opens up some
>> interesting possibilities.  Are you doing any reasoning over these
>> categories, or do you plan to do so?  Also, are there any plans to use OWL
>> to define the Kepler data type hierarchy?  Will the simpler method of
>> creating actor categories be built on top of this OWL ontology or are you
>> planning to remove the ontology and replacing it with something else?
>>
>> One more question -- do you have a tentative release date for version 2.0?
>>
>> Thanks!
>>
>> -- Ken
>>
>> On Tue, Jun 2, 2009 at 1:46 PM, Chad Berkley <berkley at nceas.ucsb.edu<mailto:
>> berkley at nceas.ucsb.edu>> wrote:
>>
>>    Hi Ken,
>>
>>    See my responses below.
>>
>>
>>    Kenneth Jones wrote:
>>
>>        I would like to build a custom version of Kepler.  I have used
>>        the Kepler build system (which is really great, btw) to download
>>        the latest version from the svn repository and I've followed
>>        your tutorial and successfully built the "Hello World" actor.  I
>>        now have four packages in my workspace:
>>
>>        hello-world
>>        kepler-1.0
>>        kepler-1.0-jar-tag
>>        loader-for-kepler-1.0
>>
>>        Everything seems to work just fine except for one thing -- the
>>        actors don't display icons.  They just show up as boxes with
>>        black borders and white interiors.  Any idea what I did wrong
>>        and how to go about fixing it?
>>
>>
>>    Are you talking about the helloWorld actor or all of the actors in
>>    the actor library?  You can set icons in the file
>>
>>  kepler-1.0/kepler-1.0-jar-tag/configs/ptolemy/configs/kepler/uiSVGIconMappingsByClass.properties
>>
>>    You just need to provide the class name and the SVG icon you want to
>>    use.  You can also assign icons by LSID (if you have given your
>>    actor a unique LSID) with the file
>>    uiSVGIconMappingsByLSID.properties in the same directory.
>>
>>
>>
>>
>>        One more question: I only want to display a subset of the Kepler
>>        actors (and some others that I will create) and organize them in
>>        a taxonomy that is different from the default Kepler taxonomy.
>>         How do I go about creating a custom taxonomy, filled with
>>        custom sets of actors and display it in the Components tab of
>>        the Components and Data View of Vergil?
>>
>>
>>    We're currently in the process of making this easier, however it is
>>    currently not an easy thing to do (especially in Kepler 1.0).  The
>>    actor library is laid out using an ontology of the actors.  You can
>>    find the current ontologies in
>>    kepler-1.0/kepler-1.0-jar-tag/configs/ptolemy/configs/kepler/ontologies
>>
>>    The ontology itself defines the categories into which the actors are
>>    placed.  Each actor then gives itself one or more semantic types
>>    which are defined in the ontology.  When the actor library is
>>    rendered, the actors are then placed into the categories for which
>>    they have a type defined.
>>
>>    In order for you to make your own classification, you would need to
>>    create your own OWL ontology, then add semantic types to the actors
>>    you want to categorize that link into your custom ontology.
>>
>>    Like I said, we're in the process of making this easier in the
>>    future Kepler 2.0 (or whatever the version turns out to be).  Let me
>>    know if you want further details.
>>
>>    Thanks,
>>    chad
>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mercury.nceas.ucsb.edu/kepler/pipermail/kepler-dev/attachments/20090603/7dd3c554/attachment-0001.html>