Re: [gupnp] Introspection is here!

On 6/30/07, Jorn Baayen <jorn@openedhand.com> wrote:
> Tervist :)

   Now that my gf wasn't able to translate but i can guess :)

> Ah OK, makes sense. I'll put it back tomorrow. A reference would still
> be appreciated if you have the book handy at some point. :)

   Chapter 6 (Desccription), page 98, the last sentence before the
title "Basic Device Information" , good that i had hlighted it. :)

> > I did read the ControlPoint code for that but i thought that is
> > only valid for asynchronous GETs while we do a synchronous GET since
> > we contruct the introspection object on demand.
>
> Ah of course, it's sync. We should provide an asynchronous version as
> well.

   Yeah! wouldn't hurt. :)

> >
> > > - Why do the ServiceIntrospection functions return copies of internal
> > > data? Why not return pointers to the internal structures? This would
> > > save a whole bunch of mallocs.
> >
> > Yeah! makes sense i guess. I was concerned about the thread-safety
> > and stuff while writing the code but now that i think about it, all
> > this data is supposed to be read-only so need for such concerns, is
> > there?
>
> gupnp isn't thread safe anyway so it shouldn't matter. :)

  Ok. :)

> >
> > > - StateVariableInfo: Rename "gtype" field to "type"?
> >
> > It was 'type' first but then i renamed it to 'gtype' to make it very
> > clear that this type is the GType and therefore not the same as the
> > 'data_type' field in the same structure.
>
> Ok, I see your point. The thing is that elsewhere in the API GType
> arguments are named "type", so I would still propose the rename for
> consistency.

  Ah ok!

> > > - StateVariableInfo: Do we really need the "data_type" and "is_numeric"
> > > fields? This info is/should be provided by the GType field.
> >
> > Some people might be interested in the UPnP type of the field, at
> > least i would be when i write the action invocation dialog of the
> > device spy, where the 'data_type' shall be reported as type and the
> > gtype will be used internally by the app to decide what widget to
> > provide for that action argument (text-field or checkbox etc).
>
> I see your point. But as we don't expose any UPnP internals in the API
> anywhere else, I would prefer to keep this API minimal also. Can't the
> device spy just use g_type_name() and min/max value stuff to denote
> types? We don't *have* to use UPnP terminology, do we?

   I also see your point and will think about it. :)

-- 
Regards,
Zeeshan Ali
--
To unsubscribe send a mail to gupnp+unsubscribe\@o-hand.com