[gme-users] RE: [udm-users] BON2 vs. UDM

[Aditya Agrawal]

-----Original Message-----
From: Abdullah Sowayan 
Sent: Friday, April 09, 2004 3:43 PM
To: udm-users; gme-users
Subject: [udm-users] BON2 vs. UDM

 

Other than supporting current interpreters, what is the economical
justification for maintaining two separate model traversal APIs (BON2
and UDM). Is there something that BON2 offers that UDM doesn't? If so,
can't similar features be incorporated into UDM thus we'd be focusing
our effort in a single unified model traversal technology?

 

[Adi] 

This is a good point that Abdullah has raised. I don't know why the two
efforts are continuing simultaneously but I can point out the positive
and negative points of each interface as per my knowledge:-).

 

Places where UDM scores over BON

 

1.	If you want to add/delete some model elements temporarily and
then run the transformation: E.g. GME's metamodeling interface has
multiple sheets and ObjectProxy's while from the meta-interpreter point
of view it should be one single sheet with a monolithic metamodel. One
way to implement this would be a preprocessing step that converts
multiple sheets into one sheet and there second step would be the
meta-interpreter. To implement such things using BON a set of temporary
data structures are created and populated to form this information. 
2.	Merging two interpreter written for the same paradigm: This is
based on another problem I faced with BON. There were two interpreters
written for the same paradigm called "MILAN". Later on the requirements
of the project chaged such that my interpreter needed to call some
functions written in the other interpreter. This however was not
possible because both the interpreters had our own customization of the
CBuildel classes and thus the two interpreter codes could not be
compiled together.
3.	Cannot make a command line utility of the BON interpreter: BON
interface can only be used to write GME interpreters. The same code
cannot be used to produce an equivalent command line utility which
becomes very important for automation of tool chains. 
4.	Two BON interfaces cannot be used at the same time: If I want to
traverse one GME model and produce another that belongs to a different
paradigm I don't think BON allows you to create two BON based projects
at the same time. For these cases you have to use the COM interface on
the target end. I may be wrong about this though.
5.	BON by default is paradigm specific: The lack of paradigm
specific API for BON missing for some. Now I hear that it has been added
in BON2

 

UDM on the other hand can seamlessly tackle all the above mentioned
issues. UDM based data networks can be copied with one API command and
you can have any number of projects opened and manipulated at the same
time. We usually have the same code used for a GME interpreter, command
line tool and also make it a library for other interpreter to call. It
provides an easy to use paradigm specific API with complete read and
COMPLETE write support. The biggest advantage of UDM is that it is not
restricted to only the mga format. The same code can be used on a
paradigm specific gme independent XML file which can be used as a way to
interchange with other tools.

 

Places where BON scores over UDM

 

1.	Reference ports: Currently UDM is not able to provide the user
with the information about which reference port a particular connection
port is associated with.
2.	Implementation and Interface Inheritance and equivalence: The
MetaGME2UML interpreter currently doesn't not support Interface and
implementation inheritance and equivalence.

 

I agree with Abdullah that one API for traversing the models will help
every one around.

 

Thanks,

Adi

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.isis.vanderbilt.edu/pipermail/gme-users/attachments/20040412/e3606d2f/attachment.htm