Re: desktop layout patch

[Michael Toomim <toomim uclink4 berkeley edu>]

Havoc Pennington wrote:
> Michael Toomim <toomim OCF Berkeley EDU> writes:
>
> > Huh?  Flicker in the pager?  This protocol doesn't affect the pager's
> > update latency.  This protocol only affects the WM's directional
> > commands.  If you mean flicker in the wm... well directional commands
> > can't really "flicker"...
>
> It could affect visible things, no reason why not. It's sort of nice
> to avoid intermediate states, I guess, as a general principle.

Ok.  So I think I agree with this "intermediate states" point.  It makes 
me see things a little differently:

We have three axes of information here: the number of desktops, the 
number of rows, and the orientation of the rows.  The first is provided 
by the WM (right?) and the second two are provided by the pager.

Right now, the pager posts a "X, Y, Orientation" tuple, which means that 
it is effectively posting the WM's "number of desktops" information as 
well as its own two items.

This means that we have "number of desktops" being transmitted from the 
WM to the pager, and then back to the WM.  Is this the problem that you 
guys see?

If so, maybe we should go with point C) from me previous post:

"C) If we did want the pager to advertise only the number of wrapped 
rows, and not the number of columns created by those rows, why not just 
have the pager advertise only the number of wrapped rows?  I suspect 
that implementations of this protocol will choose to do one or the other 
anyway, and that the special case will just result all implementations 
maintaining the same pieces of un-exercised code."