Re: Writing a gvfs backend

[Alexander Larsson <alexl redhat com>]

On Wed, 2008-05-14 at 10:29 -0600, Boyd Timothy wrote:
> Any ETA on when you're planning to stabilize the DBus interface?
> Ideally, it'd be nice if my already-running-process would just be able
> to register a DBus interface name, and I could provide an xyz.mount
> file that gvfs' daemon would just look and see the DBus interface
> already registered in the session and could go about its own merry
> way.  Thoughts/suggestions?

Some problems with this:

* The interfaces are in flux as we change the implementation during
bugfixing, and general fixes as we learn more about how things work

* The dbus interfaces weren't really designed to be public, so they are
a bit raw in places

* We use a custom binary protocol over a pipe for file transfer (really
needed or xfer performance totally dies). Also, we use direct peer to
peer connections from the clients to the daemons, not the session bus
(although finding the daemon is done using the session bus).

That said, it could very well happen in the future.