To sum up:
Tom thinks implementing non-realtime collaboration is easier to implement than
realtime collaboration.
Linus thinks implementing realtime collaboration is easier to implement than
non-realtime collaboration.
If we have non-realtime collaboration we really don't need realtime
collaboration.
Linus thinks that if we have realtime collaboration we don't need non-realtime
collaboration. Everybody else disagrees ;-). I am slowly beginning to see the
light on this.
Roy has suggested a good enough first step for non-realtime collaboration in
that it allows partitioning the model physically and logically, pushing most of
the responsibility for maintaining consistency to the user, and without
implementing any of the complicated comparing, merging, tracking changes-tools
nor the management of a model that is partially read-only. Is this a small
enough step to set up as a goal for the 0.24 release of ArgoUML?
/Linus
> -----Original Message-----
> From: Roy Feldman [mailto:[EMAIL PROTECTED]
> Sent: den 26 juni 2006 22:31
> To: [email protected]
> Cc: OSF
> Subject: [argouml-dev] Concurrent Model Development [Formerly UML model
> and CVS.]
>
> On Thu, 2006-06-22 at 02:03 -0400, Tom Morris wrote:
> > This sounds like it's intended to address real-time concurrency, but
> > there are many times when multiple people want to work on a design,
> > but not necessarily simultaneously. Also, collaboration isn't
> > necessarily limited to two people. I think we could get a long way by
> > providing better support for non-realtime collaboration before
> > addressing the realtime piece. This includes support for partitioning
> > models physically and logically, merging models, tracking changes to
> > models, comparing models effectively, etc.
>
> I just want say that my experience with modeling on large projects is
> that you really need asynchronous "non-realtime" collaboration, which I
> think of as concurrent, as opposed to simultaneous, real-time
> development. The latter is great for brainstorming. The former, IMHO,
> is of much greater practical importance.
>
> I think this is particularly true if you are going to use a UML tool for
> more than just documentation. For example, if you are using a tool like
> AndroMDA to do "Model Driven Development", then you are inevitably going
> to need concurrent model development, just like you need support for
> concurrent software development for a conventional software project of
> any significant size.
>
> Doing Model Driven Development, the model is an integral part of the
> source of the application. As a result, you inevitably need to have
> multiple developers owning and modifying different parts of the model.
> Requiring this to be done in a purely sequential fashion is completely
> impractical.
>
> That is one the main reasons that all commercial tools which support
> some form of Model Driven Development, such as "Rational Software
> Architect", support concurrent model development.
>
> ArgoUML, combined with AndroMDA, could provide a serious open source
> alternative to commercial tools like Rational Software Architect, but
> only if ArgoUML supports some form of concurrent model development.
>
>
> >
> > For realtime collaboration, choosing the appropriate level at which to
> > distribute the application is a key design decision. The elements
> > sent to clients can be anything from blocks of pixels, to drawing
> > commands, to ArgoUML meta commands, to model repository commands and
> > events. Each level provides a different set of benefits and costs.
> > The top level (sending pixels one way and mouse events the other) is
> > probably doable today with screen sharing software using an unmodified
> > version of ArgoUML.
> >
> > Tom
> >
> > -----Original Message-----
> > From: Linus Tolke [mailto:[EMAIL PROTECTED]
> > Sent: Monday, June 19, 2006 10:21 AM
> > To: [email protected]
> > Subject: [argouml-dev] Re: [argouml-users] UML model and CVS.
> >
> >
> > Hello all developers!
> >
> > For some time I have believed that the best solution to the
> > concurrent development problem is to allow the running of
> > ArgoUML in a server/client mode. I.e. a server opening and
> > saving the model and clients that have an separate GUI that
> > connects to the running UML Model and Diagram Model.
> >
> > With MDR and eventually the diagrams' model stored in MDR
> > (Diagram Interchange?) this design feels not entirely
> > out-of-reach.
> >
> > The benefits is that there is no merging of models by some
> > other tool. All Diagrams, Tree and Model elements update
> > immediately when someone has made a change, part of this is
> > already in place since we have several GUI elements working
> > against the same models already. Mauro's work can be used as a
> > help for reserving objects to avoid race conditions (if
> > needed).
> >
> > The drawbacks are that this requires that all developers are
> > on the same network but that is not such a big deal since most
> > of us are on the Internet anyway. The developers also need to
> > coordinate on who runs the server and who does not.
> >
> > I would like that you had this possible architecture in mind
> > while pondering about the design of the different layers on
> > top of MDR and the Diagram Model. Issues are with one of the
> > developers moving a Fig in a diagram, how have the other
> > developers' clients registered interest to learn about this
> > move and thousands of similar things... Other issues are with
> > things like Undo. Should each user have their own Undo list?
> > Probably. Will that work? Not always.
> >
> > /Linus
> >
> >
> > ______________________________________________________________
> > Från: Linus Tolke [mailto:[EMAIL PROTECTED]
> > Skickat: må 2006-06-19 15:52
> > Till: [email protected]
> > Ämne: Re: [argouml-users] UML model and CVS.
> >
> > Yes, Marco!
> >
> > Your original question was how it is possible to have the
> > model within CVS and my answer is avoid concurrent
> > development.
> >
> > If you want concurrent development, then I there are several
> > different ways and reasons to achieve this.
> >
> > Mauro's paper Usando a modelagem colaborativa no aprendizado
> > da UML available from the ArgoUML web page at
> > http://argouml.tigris.org/docs/ discusses a solution where you
> > do this for the purpose of stimulating collaboration (if I
> > have understood it correctly, I don't read Portuguese). There
> > has earlier been other attempts at this where ArgoUML just
> > opens an extra window (which also can be achieved by running
> > ArgoUML in a NetMeeting/Remote Desktop environment).
> >
> > This is probably not what you want but I want to mention it as
> > an alternative for you to think about.
> >
> > /Linus
> >
> > mailto:[EMAIL PROTECTED]
> > >> Sent: den 17 juni 2006 14:39
> > >> To: [email protected]
> > >> Subject: [argouml-users] UML model and CVS.
> > >>
> > >> UML is great and ArgoUML is great too.
> > >>
> > >> But with respect to MDA I was wondering how it would be
> > possible to
> > >> develop UML model in a CVS as well as we develop today a
> > java class.
> > >> CVS can resolv conflicts on java files but not on ArgoUML
> > files
> > > because
> > >> they are binary files.
> > >>
> > >> And even if we use xmi I not sure CVS can resolv conflicts
> > if xmi
> > > export
> > >> do not take care to export objects always in the same
> > sequence.
> > >>
> > >> Any ideas about it?
> > >>
> > >> Regards,
> > >> Mar
> > >> Compilo subAdministrator
> > --
> > Ing. Marco LOMBARDO
> > =============================
> > Mayking spa
> > Via Brescia 31
> > 36040 Torri di Quartesolo (VI)
> > Cell +39 347 1979448
> > Uff +39 0444 267561
> > Fax +39 0444 269945
> > email: [EMAIL PROTECTED]
> > Skype: lombardomayking
> > web: www.mayking.com
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
