Re: (FC-Devel) UXF & XMI, expectations

[Jason Elliot Robbins <jrobbins@cadet.ics.uci.edu>]

Danny Werner wrote:
>Would you share (some of) those expected features with us, Jason?
>That might help getting the requirements more specific.

Here are some things that several people have asked for in Argo/UML
but implementing these is not in line with my research goals for the
next year:

 Reverse engineering
 C++ support
 Full Undo and Redo
 UML Templates
 Support for patterns in various ways
 Repository for designs
 Methodology-specific features
 Integration with MS Visual Studio or some other IDE or VC tools
 Reading existing rational rose files
 Many of the finer points of UML state-diagrams and sequence diagrams

I don't know which commercial tools provide each of these, probably
none provide all of them, but interoperation will ideally allow users
to use whichever commercial tools they have.  I expect that commercial
tools will start supporting XMI in about 6-12 months.  There will
probably be a revised XMI to match UML 1.4 in April.



>> I chose to use PGML to represent diagrams rather than XMI.
>
>Why? 

PGML (Percision Graphics Markup Language) is a proposed web standard
for viewing "vector" graphics (lines, ciricles, polygons, images,
etc.).  It is done in XML.  The PGML standard includes specifi tags
that cover pretty much all of the graphical elements used in UML.  In
contrast the UML and XMI standards leave the details of ViewElements
unspecified.  Also, assuming that PGML becomes a widely used standard,
off-the-shelf web browswers should be able to view PGML diagrams, thus
allowing people to easily look at work done in Argo/UML even if they
don't have the acutal tool.

>Does this mean that you think FreeCASE should support XMI
>early on?

I think FreeCASE needs to take the path-of-least-resistence and focus
on short-term goals until it has something running that looks a little
like a real CASE tool.  XMI is a good long term-goal, but without
something in the short-term FreeCASE is going to have a hard time
motivating contributors.  Maybe a subset of XMI is the way to go.  In
Argo/UML I first used Java ObjectSerialization because it was the
easiest thing to do; however, after 6 months I found I had to switch
because of difficulties with inter-version file compatability.  Now I
am working on support for about 70% of XMI, with some of the remaining
30% to be done next month.

jason!