xc/programs considered harmful

[Kevin E Martin]

On Fri, Dec 17, 2004 at 06:44:36PM +0100, Daniel Stone wrote:
> On Fri, 2004-12-17 at 09:36 -0800, Stuart Kreitman wrote:
> > Daniel Stone wrote:
> > >I would be happy to formally propose the move of these to the modular
> > >tree.
> > 
> > The modular tree methodology is not clear- is there a roadmap
> > for the transition?  What does it mean to move sources to the modular
> > tree? Let's just say there will be a 6.8.3 before modular is ready.
> > Would xterm Not ship in 6.8.3?
> 
> Presuming there was a 6.8.3, then that would obviously have to come off
> XORG-6_8-branch, which would still contain xc/programs/xterm.
> 
> Here's the roadmap:
> 	* Start moving stuff from the monolithic tree to the modular tree.
>     * Declare the move finished when there is nothing left.
> 
> By 'move sources', I mean just that.  You remove the sources from the
> monolithic tree (leaving history intact, of course), and then you add
> them to the modular tree, again with history intact (either by a
> straight cp -R if you really want to, or via a CVS vendor branch if you
> care about better preserving line-of-development visibility).

Many of us have talked about this move to a modular tree but most are
currently caught up in the 6.8.2 release.  I too would like to see us
move forward on the modularization front, but I don't think that
removing bits from the monolithic tree is the right answer to get us
moving forward.

There will still be people that need to maintain the old monolithic
tree, and I think it is fine for them to do so.  The main reason for
this is that there will be vendors that will take time to move their
internal X development and releases over to a modular system -- not
everyone can switch as fast as you or I can.  And, removing useful bits
from that tree would cause significant problems for those vendors that
are not ready to move.  I think a better proposal would be to map out a
transition plan, keep the old monolithic tree around until the
transition is complete, and then deprecate it after we've shown that a
modular release is just as stable and reliable as the old monolithic
ones (or hopefully much more stable with better features).

I would like to see a concerted effort to define exactly how the
transition will occur, but I agree with others that perhaps it is going
to be difficult to get consensus right now since many people are about
to leave for the holidays and some already have.  There are several
proposals that have been floating around, which should be discussed.
Keith Packard and I (and others) talked about one at a recent meeting,
which I will try to write up over the weekend so that we can discuss it
here.

Kevin