[daniel at fooishbar.org: Re: [Bug 17372] New: New user account for cjb, Chris Ball.]

[Adam Jackson]

On Sun, 2008-08-31 at 16:36 -0300, Tiago Vignatti wrote:
> Daniel Stone escreveu:
> > For those who care about KDrive ...
> > 
> > ajax did a pretty good job in removing all the DDXes and unused APIs like
> > KAA, but I think there's a pretty long way to go until we have something
> > that's really ideal.
> > 
> > Of course, I'd desperately love to see it unexist, but hw/xfree86 is
> > still a little while away from being that server.  Our tablets will be
> > running Xorg shortly, but hey, we're talking about 256MB+ RAM, 256+ MB
> > flash, and an ARM11, so it's more MID than tablet/phone/consumer device.
> > 
> > So it's probably worth caring about for a little while yet: even if
> > hw/kdrive does just become hw/xephyr, there's still a fair bit of work
> > to do there.
> 
> cool, this is pretty interesting Daniel. And what about Xnest?

>From a coverage perspective, it's very pleasant to have the two sorts of
nested X server, one that's a protocol repeater and one that's a
software rasteriser.  However, we have two protocol repeaters: Xnest,
and Xdmx.  Ideally what I'd like to see is Xnest implemented as an Xdmx
personality; you should be able to just symlink the one to the other and
have it act like the Xnest of old.

>From a non-coverage perspective, the Xephyr approach to nesting does not
scale.  Xdmx had a xephyrish mode once, and it just doesn't perform well
beyond a very small number of displays.  There are any number of reasons
for this - lack of compressed image transport for example - but it does
indicate that both approaches could be worth keeping for reasons of user
experience as well as pure test coverage.

But in the long term, yes, kdrive needs to go away.

- ajax