gtg-contributors team mailing list archive

[Karlo Jež <karlo.jez@xxxxxxxxx>]

On Fri, Jun 11, 2010 at 4:22 PM, Paul Natsuo Kishimoto <
mail@xxxxxxxxxxxxxxxxxxx> wrote:

> On Fri, 2010-06-11 at 13:24 +0200, Bertrand Rousseau wrote:
> > Le 11/06/10 13:18, Luca Invernizzi a écrit :
> > > On Fri, Jun 11, 2010 at 9:53 AM, Lionel Dricot<ploum@xxxxxxxxx>
>  wrote:
> > >> Hi all,
> > >>
> > >> Today I've merged Paul's code reorganisation.
>
>   Only the first of many parts! Thanks :)
>
>  Next up is the CLI.
>
> > >> I've one comment :
> > >>
> > >> - the dbus was moved into GTK. I believe that it should not. Instead,
> > >> there should still be a "view manager" that would be UI agnostic and
> > >> contains the DBus interface.
> > >>
> > >> Any UI, should have to register itself to that view manager and
> implement
> > >> a given interface so that, when the viewmanager says "open task X", it
> > >> does.
> > >>
> > >>
> > >> What do you think ?
> > > I agree. We could have a view-manager for each UI, and a
> > > meta-view-manager that just acts as a proxy to the current VM.
> > > Registering a ui means registering its VM to the meta-VM.
> >
> > Just my 2 cents...
> >
> > Is it me or the "viewmanager" is actually the controler in an MVC model
> > perspective? In this case, couldn't we just name this file after this
> > naming scheme? It would be more directly identifiable.
>
>   That was how I thought of it, too. If we rename, I would just call it
> GtkUI and put it in GTG/gtk/__init__.py.
>
>  At risk of stating the obvious, the extent to which we want to provide
> shared code that is common to *all* UIs depends on what those UIs *are*.
> Currently:
>
>      * GTK
>      * Command-line
>      * Web
>      * (Could there be more? Maybe a Qt UI?)
>
>  Anyway, the three are very different. For the example "the ViewManager
> says 'Open Task X'," I can't imagine how that would be useful for the
> CLI or Web interface.
>
>  Also, once the client-server split is accomplished, if actions against
> the server are atomic, then (in theory) all three UIs should be able to
> access the server simultaneously without anyone (i.e. a MetaViewManager)
> needing to keep track of them.
>
>  We already realized that the current DBus interface will need to be
> split in at least two, and that will indeed happen. One part for the
> server. Maybe the second part should be for *GTK* UI specifically,
> instead of "the UI" abstractly.
>

I agree with Paul. With a properly done (atomic) dbus interface, there
should be no need for a server-side meta view manager, or any registration
of the interfaces. I guess we could do it if we wanted to give higher
privileges to the UIs, as opposed to the other applications that access gtg
tasks using dbus (i.e. docky), but I don't see the need for that.


> --
> Paul Kishimoto
> MASc candidate (2010), Flight Systems & Control Group
> University of Toronto Institute for Aerospace Studies (UTIAS)
>
> http://paul.kishimoto.name — +19053029315
>
> _______________________________________________
> Mailing list: https://launchpad.net/~gtg-contributors<https://launchpad.net/%7Egtg-contributors>
> Post to     : gtg-contributors@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~gtg-contributors<https://launchpad.net/%7Egtg-contributors>
> More help   : https://help.launchpad.net/ListHelp
>
>