Re: interesting article on XFree86 speed

<[email protected]> Mon, 28 Oct 2002 07:05:15 -0800 (PST)
Newsgroups gmane.comp.xfree86.render
Message-ID <[email protected]>
Sorry if this is a mostly a repeat: my mailer hiccupped.

Yes, 20ms is an eternity, unless you are going long haul or via low bandwith
lingks.

I think there is alot going wrong here:
	1) stupid toolkits
	2) stupid apps that do unnecessary repaints.  I've often detected
apps repainting N times where one would do (try running X on a dialup line
sometime: it is enlightening to see how stupid some apps/toolkits have
been).
	3) there are some places where the X protocol should be fixed.
See Hania Gajewska, Mark Mark Manasse's, and Joel McCormack's paper
at http://www.std.org/msm/common/WhyX.pdf.

OK, there are a number of things we can/should do about it:
	1) Keith has managed to get the X protocol dumper out from
	under restrictive license.  There are, however, a bunch
	of versions to be reconciled: help here would be appreciated,
	to make this tool widely usable by lots of people.
	2) fix broken toolkits and apps.
	3) Keith and I chatted with Owen about what to do with his
small extension to fix embedding, and have come to the conclusion we
need a "fix the protocol" extension, to fix items that matter (some
of what Hania/Mark/Joel complained about probably don't matter in
current contexts).  Names for this effort greatly appreciated
(otherwise we'll just call it the "Norm" extension: see if you can
figure out the etimology of that name).
				- Jim


--
Jim Gettys
Cambridge Research Laboratory
HP Labs, Hewlett-Packard Company
[email protected]