Re: RFC: Adding zlib dependency to libgio

[nf2 <nf2 email gmail com>]

On Mon, Nov 9, 2009 at 4:05 PM, Thiago Macieira <thiago kde org> wrote:
>
> Besides, glib is only a dependency of Qt on the X11 platform. I can justify a
> VFS API that requires D-Bus to work properly (with some effort, on some
> platforms other IPC mechanisms would be preferable), but I cannot do it if it
> requires using glib & gobject in platforms that its own maintainers currently
> don't support.

Hmm ... I think a VFS API shouldn't even require D-Bus to work. Just
like GIO only uses D-Bus/GVFS on X11. I agree that a QtVFS shouldn't
depend on GIO in the API. But perhaps it would make sense to design it
close to GIO so that it can act as a thin binding when GIO is
available.

My feeling is that the tough part of a new QtVFS is designing and
consolidating the API. Therefore my thought was: If GIO has proven to
be portable, an API which is designed to wrap GIO should be portable
itself. Without necessarly depending on GIO. So this approach might
save lot's of time. Cloning GIO in QObject/C++ style should save
months racking one's brain about little details like which virtual
methods and signals are required in which class.

And if such an API moves towards a pure QDBus implementation later,
why not. But starting with that might be putting the cart before the
horse. Because I'm a bit worried that a complete rewrite of GVFS in
Qt/C++ is just a too ambitios undertaking to get started.

Regards,
Norbert