[RFC] removal of server generation mechanism

[Adam Jackson]

On Sun, 2009-10-25 at 13:47 -0700, Keith Packard wrote:
> Excerpts from Tiago Vignatti's message of Sun Oct 25 12:31:24 -0700 2009:
> 
> > Therefore, I'd like to see a strong argument to why _not_ remove such 
> > regeneration code.
> 
> Server reset is defined in the X protocol specification; removing this
> would make the server intentionally non-conformant.

This is pedantically true.  Excepting a few details, you could get the
same result by just re-execve()ing yourself.  There's nothing in the
protocol spec saying the server needs to retain the same PID across
resets (and indeed, for remote connections, you couldn't know).

I think the only real detail that would need to be handled would be
passing the listener sockets to the new process, since otherwise some
other process could grab them while you're resetting.  Pretty sure even
Windows has a mechanism for that.

> Furthermore, for display managers, 'reset' provides a strong guarantee
> that a subsequent user session will not be affected by applications
> running in the current session.

execve() is an equivalent guarantee.

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
Url : http://lists.x.org/archives/xorg-devel/attachments/20091026/9c0ed31b/attachment.pgp