xc/programs considered harmful

[Adam Jackson]

On Friday 17 December 2004 11:34, Roland Mainz wrote:
> Daniel Stone wrote:
> > Hi all,
> > Does anyone realistically care about[0] the following directories:
> > xc/programs/xterm
>
> xterm is being used and should stay exactly at that place. Even when the
> only reason is to avoid screwing-up the CVSblame.

You do know that you can move directories within CVS without losing history, 
right?  All that matters is the ,v files.

xterm has a maintainer, and it's not us.  You don't need it to build anything 
else in our tree.  Why is it in our tree?

> > If no-one objects (read: steps in and starts maintaining the relevant
> > program) to the removal of any of these, I would like to kick them from
> > the tree ASAP
>
> Daniel: Why is that neccesary ? The only "gain" here (beyond to push
> your pet project "modular tree") is to cause trouble for other people,
> nothing else.

If "getting the bits you want from a different package than you're used to" 
counts as "causing trouble", then:

- every release causes trouble
- switching from XFree86 to Xorg caused trouble
- the Xprint merge caused trouble
- the Cygwin merge caused trouble

Daniel was not suggesting deleting those apps; only moving them from the 
monolith to the modular build environment.

The idea here is that if there's a bug in some underutilized application like, 
say, xedit, and you actually care about xedit, then you can download the new  
xedit release from xapps and build just that piece and move on with your 
life.  This gets working software to the users faster, without the need to 
track CVS.

What a radical idea.

> > They have no place in the tree as we move towards a
> > monolithic structure, and at least one (xterm) has a very active
> > upstream.  If xterm remains in the monolithic tree, it should be moved
> > to extras/,
>
> It may be better think about doing the xterm development in the Xorg
> CVS.

Thomas Dickey (xterm's maintainer) is not interested in moving his development 
hosting.  Feel free to attempt to convince him otherwise, but I doubt you'll 
succeed.

Since he's not going to move into Xorg CVS, all we'd be doing by applying our 
own changes to our own copy is diverging the source trees and creating more 
merge hassle.  Merges suck.  Let's not do them anymore.

People _have_ proposed modular breakouts for xapps and xlibs:

http://freedesktop.org/wiki/Software/xapps
http://freedesktop.org/wiki/Software/xlibs

And the modularisation we did in debrix for the server is, basically: server 
in one component, one component for each loadable object in the server 
besides the int10/scanpci/ddc/vbe junkpile.  Surprise, it worked.  It was an 
absolute joy to work with too.  Rebuild the server in 9 minutes on my slow 
4.5-year old workstation.

Your refusal to consider these as serious proposals simply because their 
discussion hasn't taken place on xorg_arch is, IMO at least, tantamount to 
putting your head in the sand.  The @x.org lists are anachronistic, all the 
real work is happening here and on r-w at .  And the @x.org lists are still 
broken:

http://lists.freedesktop.org/pipermail/xorg/2004-October/004055.html

So please.  Come up with serious technical objections to modularising the 
world.  That way we can address them and _fix_ them, rather than continuing 
to fight the monolith for the rest of our lives.  Right now all I'm hearing 
from your corner is Fear, Uncertainty, and Doubt.

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.x.org/archives/xorg/attachments/20041217/491ab129/attachment.pgp>