xc/programs considered harmful

[Daniel Stone]

On Fri, 2004-12-17 at 15:03 -0500, Owen Taylor wrote:
> What really disappoints me here is that modularization has been 
> discussed, experimented with, tried out for several years, and nobody
> has sat down and wrote down a concrete plan for:
> 
>  - How will the code be structured in CVS

http://cvs.freedesktop.org/xlibs/
http://cvs.freedesktop.org/xapps/

And Debrix, which has one module for the server at large (server +
loader + 'core' modules [e.g. scanpci, vbe, int10, whatever]).  It's not
up right now, and I don't have the time/resources to do it for a few
days yet.

>  - What will be the released tarballs

... tarballs ... of code ...

>  - What are the stages for moving code to match 

Within CVS, either do an import from a vendor branch, or do as we have
already done (the modular libraries are there and working, dude) and
just cp the ,v files.

> If I hadn't talked to people about it in person, I wouldn't believe
> that people even had a plan.

Well, I'm sure the existence of a working codebase will reassure you
otherwise.

> Such a plan could be debated upon, refined and agreed upon. I don't
> think "if nobody objects I'm going to start moving things" is a 
> decent way of making this kind of decision. For one thing, *some*
> people will (as we've seen on this thread) object, even if there
> is a general consensus.

It seems there have been some rather vehement objections, yes.

But I do not see the difference, at any fundamental level, between what
you have proposed, and what was done.  Ho hum.

Had I been aware of how full this would make my inbox, I would've made
the subject a little less flippant, I suppose.

> It's great if development can proceed by consensus, rather than 
> having to be voted upon. But I don't see how there could be a 
> consensus on something that's just in the head of a small subset
> of people. Proceeding to make changes before documenting and 
> explaining the plan is just inviting confusion, flamage, and
> motion in circles.

I do not see that any changes have been made.

What exactly would you require?  A ten-point transition plan?  The plan
for moving to the modular tree has been rather clearly laid out:
	* Break the tree up into logical components -- DONE.
	* Prove in an external repository that the concept is workable -- DONE.
	* Start moving parts of the tree out to external repositories and
deprecate their monolithic component, unless they are completely dead
code, in which case they should be left to their fate[1].
	* When 3) is complete, there will be nothing left in the monolithic
tree.

Had I had any idea of the flamage this was going to generate, I probably
would've phrased it a bit better, sent out a more verbose email.  As it
is, my connectivity is not great[0], so I won't be able to actively
participate in this thread much; I fly back home on the 22nd, arrive the
24th, and I'd like to take the opportunity offered by a fortnight-odd of
holidays to catch up on sleep that I have sorely missed lately on
account of work and fd.o.

-d

[0]: I checked out of my hotel in Mataró this morning, I'm sitting in an
internet cafe in Barcelona near the beach writing this, and I'm going to
see if I can't book me on a flight to France this evening.  So,
'unspectacular' is a good way to describe my connectivity for the next
few days.
[1]: If they are important, surely someone will jump up and offer to
maintain them.  If it can't sustain external maintainership, that tells
you something pretty crucial about the codebase.