Re: KDE Interop [Was: D-BUS background]

[Havoc Pennington <hp redhat com>]

On Wed, Mar 05, 2003 at 12:54:09AM +0100, Rodrigo Moya wrote: 
> I put my vote for sharing as much with KDE (and others) as possible, and
> I really like (and follow) the discussions taking place in
> freedesktop.org, but if sharing means replacing all GNOME technology
> "because it uses gobject", then I think all the concerns raised by
> people on this issue are totally right. If they are to be replaced with
> technically better implementations, then that's ok, but based on poor
> arguments such as "it uses gobject" might be a good reason for KDE, but
> not at all for GNOME.

To really be able to share high-level implementation we're going to
have to have a common shared runtime/component system, as in the
infamous "Hub" papers. Not going to happen very quickly.

However, the fact that sharing *most* implementation isn't practical
for now doesn't mean we can't share some key high bang-for-the-buck
items.  And there are also some good high bang-for-the-buck items that
we can do by sharing specs instead of implementation.

As an aside, we should distinguish being unwilling to *link* to a
GObject/QObject API in the platform, from being unwilling to
communicate with an external process that is implemented using
GObject/QObject. Not sure people care about that distinction always,
but I think the distinction can be important.

But the basic point is, yes right now we have to weigh the benefits of
shared implementation vs. the benefits of using a decent high-level
runtime system to write the implementation; hopefully we can change
that in the star trek future, in the meantime shared implementations
will have to bring some level of value (bang-for-the-buck) before
their benefit outweighs the cost.

Havoc