byorgey:

>    What's the current status on this?  Don? Spencer?

Currently Spencer's out of action, and expressed some doubts, which is
making it hard to reach a consensus here.

>    Summary so far:
>
>      * Andrea's patch moves runLayout into the LayoutClass, changing the
>    signature slightly so that layouts receive all possible information.
>      * This would necessitate updating all *users* of LayoutClass to use
>    runLayout instead of doLayout/emptyLayout.  The patches for these changes
>    are already available.
>      * New instances of LayoutClass can choose whether to implement
>    runLayout, doLayout, or pureLayout, and so on.  Existing instances of
>    LayoutClass will not be affected (except those that are also users of it,
>    e.g. LayoutModifier).
>      * This provides a nice solution (IMO the best) for PerWorkspace.

>      Also:
>
>      * If this patch is applied to the core, I am willing to take
>    responsibility for making sure the contrib library is updated
>    appropriately, and that thorough documentation is made available to guide
>    those wishing to understand the architecture of the LayoutClass.
>      * David and I think this is elegant and should be applied.  (See David's
>    prior e-mail for an excellent discussion of some of the issues involved.)
>      * Spencer has voiced discomfort with the style of this change to the
>    API.
>      * Don has not expressed an opinion one way or the other.

I'm happy if Brent and David are on this one.

>    It would be nice if a decision could be reached on this one way or the
>    other, so we can move ahead with PerWorkspace and other things.  I'm not
>    trying to be pushy, just trying to spur things forward in a friendly way.

As Spencer only expressed doubt here, and I'm persuaded by Brent's
arguments (and his offering to champion them), let's proceed.