Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!purdue!decwrl!decvax!crltrx!max.crl.dec.com!jg
From: jg@max.crl.dec.com (Jim Gettys)
Newsgroups: comp.windows.x
Subject: Re: Xserver wanted which supports multiple prioritized threads
Message-ID: <137@crltrx.crl.dec.com>
Date: 8 May 89 14:31:19 GMT
References: <396@isi.UUCP>
Sender: news@crltrx.crl.dec.com
Organization: DEC Cambridge Research Laboratory
Lines: 18

Input is normally sent to clients as soon as possible; good X implementations
check for input with great frequency (poor ones may not, so your milage may
vary), by use of shared memory to make the input check cheap.  Of course, you
may not see this for a while if other clients are keeping your server busy,
as the client who gets the input has to do something about it and it may be
a while before its output is displayed.  But a quality implementation will
process input between each request from any client (i.e. as fast as possible).

As to your first request, it would require a server extension to express
the priority, and a more complex (read slower) dispatcher to implement it.
No one to my knowledge has done what you are asking for.  Why don't you
help define an extension?  Note that you may have to think about what happens
when things are grabbed, to prevent possible deadlock situtations.

Note that finer than single request grain scheduling cannot be done in X;
otherwise graphics output cannot be guaranteed to be correct.  (and I shudder
even thinking about implementing such a system).
			- Jim Gettys