ubuntu-appstore-developers team mailing list archive

[Colin Watson <cjwatson@xxxxxxxxxx>]

On Fri, Jul 12, 2013 at 04:42:25PM -0500, Ted Gould wrote:
> On Fri, 2013-07-12 at 22:37 +0100, Colin Watson wrote:
> > On Fri, Jul 12, 2013 at 03:31:10PM -0500, Ted Gould wrote:
> > > I'm going to suggest that we add an entry to the root item called
> > > "applications" that contains and associative array.  The index to that
> > > array is the name of the application and the value is an object of
> > > properties describing it.  Only one of which I'm suggesting today,
> > > "type" with an only valid value of "desktop".
> > 
> > Since desktop files are going to need to be associated with hooks
> > anyway, and attaching to hooks requires entries in the manifest, please
> > hold off on defining anything like this until we've got the hooks
> > actually working in practice.  It may turn out to be redundant.
> 
> I guess the question I'm being asked is "when can I start a Click
> package?"  And to do that I need to know that the "camera-app.desktop"
> file is the one to use for "com.ubuntu.apps.camera".

Right, but what I'm saying is that if we need a hooks manifest subtree
anyway then you might as well walk that rather than duplicating it in
applications.  I'm conscious that the manifest file is supposed to be
constructable by developers by hand, even if many of them do it via
QtCreator, and if possible I want to avoid making it error-prone by
duplicating information.

> When do you expect to have the hooks working in practice?

I don't think I can do it without examples to work on; that's just how
my brain works.  So it would really help if you could give me your side
of the hook: as close as possible to a script to take a desktop file as
we'd find in an app and install it in the desired way in a real Ubuntu
Touch system so that it will appear as a launchable app.  If it has gaps
where you need to know about some feature of click before you can write
them, that's fine, but I need to know the kinds of parameters I need to
support in hooks before I can implement them properly, and the ability
to experiment would be very helpful.

(I've said much the same thing to Steve Beattie, but in case the message
didn't get across: I would prefer to be given a partial implementation
of the apparmor confinement hook which I can experiment with and use to
flesh out the hooks design/implementation, than for the security team to
struggle to try to fit it into the current rough draft.)

-- 
Colin Watson                                       [cjwatson@xxxxxxxxxx]