[Uml-devel] Thoughts on Umbrello 1.3

Harald Wallus wallus at results-hannover.de
Tue Jan 20 09:42:02 UTC 2004

Dear Andrew,

MDA (model driven architecture) has one more fact:

I can use in my uml model a complex structure.
E.g. I call an object stereotype <<entity>>, and my mda-modeller will 
create a full structure to save this object into db.
The same idea is, what in java is a bean for. A bean is complete and can 
run on its own.
With get/set-methods it is possible to do modify that bean. Events 
allows, that this bean write into db or read from it,
or calls other part of application.
The wish of mda is, to start from a graphical workflow, which is 
realized into uml, from which is create  guis and db-Interfaces.
Programmer have then only to do some special work ( to program what is 
brutto of something and so on, but this can be include into  a 
stereotype to).
That is what enterprise java beans is designed for.


Next week I will find ime to test umbrello once again.

Andrew Sutton wrote:

>>Let's see. I haven't seen a release plan, but KDE 3.3 or 4.0 will need
>>another year to be released. UML2 is scheduled, I think, for this summer.
>>This means we have enough time to go for UML2. What are your concerns
>>against UML 2? I think it is really valuable to have things like MDA.
>for those copied on this that don't know what mda (model driven architecture) 
>is, here's a brief description:
>the omg (people that make uml) want to build all software using uml models. 
>the idea is that you build a platfrom independant model (classes, behaviors, 
>etc) and then transform that to a platform specific model (psm) based on 
>specific rules for the mapping. for example, you can build a pim of your 
>software in uml and then use the transformations to create models specific 
>to... i don't know Java, MS .Net, Mono, C++ on Linux with Debian, C++ no 
>Windows, Solaris, Mac... whatever the the specifics of the platform are...
>why do i propose waiting on uml2? a couple reasons
>	a) it's not done yet. we need to wait for the xmi to mof mapping before you 
>can really start implementing it. this is the major reason.
>	b) i've spent the last year working on the 1.x series implementation and i'm 
>not ready to focus on the specifics of uml2 yet. because i'll basically end 
>up completely re-evaluating the type system and having to rewrite the entire 
>thing from scratch. moreover, uml2 is way different than uml1.x.
>	c) it sucks implementing uml in c++. python would be so much easier :)
>also, you can do MDA in uml1.x. it's basically the same process. uml2 is a 
>complete redesign of the metamodel intended for extensibility. it's harder to 
>extend uml1.x.
>eventually, i'll probably start work on 2, but not until my 1.x implementation 
>is done.
>The SF.Net email is sponsored by EclipseCon 2004
>Premiere Conference on Open Tools Development and Integration
>See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
>Uml-devel mailing list
>umbrello-devel at kde.org

Dr. Harald Wallus
Results GmbH 
Am Listholze 78, D-30177 Hannover 
Tel: +49(0)511 90951-23  Fax: +49(0)511 90951-90 
Email: wallus at results-hannover.de 
Internet: http://www.results-hannover.de  

More information about the umbrello-devel mailing list