[patch] abstract socket support for xtrans

[Jamey Sharp]

On Tue, Mar 21, 2006 at 04:56:53PM -0800, Alan Coopersmith wrote:
> Jamey Sharp wrote:
> >I think there would be few objections to eliminating xtrans
> >altogether, and perhaps much rejoicing. :-)
> 
> The only transport we support on Solaris in xtrans other than sockets
> is named pipes, which once upon a time were faster than Unix sockets
> on Solaris, but I don't think are today.   (Our shared memory
> transport is actually done in libX11 and the Xserver itself, and as
> far as xtrans is concerned is just a Unix socket or named pipe that
> doesn't get a lot of data written over it.)

Is this something I should know about for XCB?

> It was nice to have a central place to do much of the IPv6 updating,
> but it didn't cover enough to make it the only place I had to touch.
> 
> My biggest question about removing it is are we sure that sockets are
> the one and only transport API we will ever want to use, on all
> systems, for all time?

No, of course we aren't sure. :-) But I'm pretty sure that xtrans is
*not* the transport API we want to use on all systems and for all
time...

The xtrans abstraction doesn't seem to be buying us anything today. As
you and the original poster have both found, it doesn't abstract enough
to keep transport knowledge out of the callers; and ISTR its interface
isn't a great fit for the available operations. If an abstraction layer
would help at some point in the future, we can design one that suits our
requirements then.

--Jamey
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 191 bytes
Desc: Digital signature
URL: <http://lists.x.org/archives/xorg/attachments/20060321/ed62fa35/attachment.pgp>