Re: Our (real) problems

[Michael Meeks <michael ximian com>]

Hi Dan,

On 4 Sep 2001, Dan Winship wrote:
> Why, is the "broken" signal tied to X?

        It ain't :-) it uses the unix domain socket, and listens for
the connection being destroyed.

        Using the X socket is problematic with regard to re-parenting,
but apparently we can work around this by doing re-parenting more
cunningly.

	Of course, freeing an object when all it's open connections
are torn down has some interesting lifecycle issues too :-) but it
seems a good enough solution for the control / control-frame pair to
me.

	Since we're interested only particularly in local machines,
for control / control-frame cases we can rely on the somewhat 'nicer'
more predictable and rapid behavior of UDSs vs. general TCP timeouts I
think, and avoid pinging.

        Regards,

                Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot