Re: NET WM Spec implementation and changes

[jg pa dec com (Jim Gettys)]

It isn't all that hard to do your own Xlib routine, even those implementing 
batching. Implementing Xlib routines is done by people implementing 
extensions, after all.  I suggest, however, that if you'd like it ubiquitous 
in the X community, you recode it into banal C; not everyone will buy 
into the C++ religion...  Avoiding round trips by batching isn't very 
hard.  To remind people, I enclose my previous pointer to where to find 
a good example of avoiding round trips implemented in Xlib.

I recommend sending your revised spec out for general review when
you've a nearly "final" spec, probably to the XFree86 developer list.
				- Jim



From: jg@pa.dec.com (Jim Gettys)
Resent-From: wm-spec-list@gnome.org
Date: Tue, 11 Jan 2000 09:32:48 -0800 (PST)
To: Matthias Ettrich <ettrich@troll.no>
Cc: wm-spec-list@gnome.org
Subject: Re: status?
-----
> From: Matthias Ettrich <ettrich@troll.no>
> Resent-From: wm-spec-list@gnome.org
> Date: Sun, 9 Jan 2000 20:30:18 +0100
> To: wm-spec-list@gnome.org
> Subject: Re: status?
> -----
> On Sun, 09 Jan 2000, Havoc Pennington wrote:
> > Hi,
> >
> > Is there a feeling that people are happy with the spec? Any final
> > additions? I noticed a few comments in the archives since the last
> > draft, but not too much traffic.
>
> Did we decide what to do with the property packing?
> Originally we intented to pack everything regarding client windows into one
> big,
> extensible property to avoid roundtrips to the server.
>
> Somebody mentioned that it's just a few lines X-lib hacking to be able to
read
> several property in one roundtrip, so we better keep it simple.
>

It was me...  It isn't all that hard to write Xlib routines to "do the
right thing" in a single round trip (you might claim I'm biased, having
written Xlib in the first place).  I have evidence on my side, however,
there having been significant X extensions written over the years.

There are things I'd do differently if I had them to do over again, so
handling errors is more of a problem than it might be, but it isn't all
that difficult.

Possibly the best example of doing this sort of batching is
in the routine XInternAtoms, in the file xc/lib/X11/IntAtom.c in an
X distribution: interning atoms became a Xt toolkit bottleneck, and so
this routine was added (and Xt reworked) to let you Intern a bunch of
atoms in one round trip.  XInternAtoms is between one or two pages of code.

                                - Jim

--
Jim Gettys
Technology and Corporate Development
Compaq Computer Corporation
jg@pa.dec.com


--
To unsubscribe: mail wm-spec-list-request@gnome.org with "unsubscribe"
as the Subject.