On Jul 7, 2008, at 1:50 PM, Martin Costabel wrote: I know you are not reachable by rational argument; if I forget it, I just have to run "ls -l /usr/lib/libreadline.dylib" ;-)
Now now Martin, no need to get carried away here. I am certainly "reachable by rational argument" and have changed my mind on issues like this quite number of times when it was clear I was wrong. I also have more than a few scars from fighting "user request" battles where no one else inside Apple apparently agreed with me, so it's not like I'm always inclined to take the easy road, either.
Again, as Jeremy has explained a great number of times now, the xmkmf bits have been abandoned by their own maintainers in favor of the autotools (Which I personally hate, but the user community argued successfully that it didn't matter whether I hated them or not. See? There's a fine example of a battle I lost to "rational argument.")
If anyone seems immune to rational argument, it might be you here given that you bring up libreadline.dylib despite the fact that it has been pointed out many times that GNU readline's license makes it completely unsuitable for anything but the most rigidly controlled "GPL only" environments - even the Linux folks have troubles with it: http://lists.debian.org/debian-devel/2002/05/msg00192.html. Until someone is successful where everyone else appears to have failed in arguing for the LGPL, as is used by every other popular GNU library, the only real alternative is to help the libedit folks to continue their work in making it fully libreadline compatible (I have also offered to help sponsor this work and got it listed as one of the GSoC projects last year, but so far progress has been slow). There was absolutely nothing "irrational" about the decision you refer to - it was quite necessary in order to provide some basic "readline" functionality without also creating a potential legal problem for anyone accidentally linking with it (as our much beloved autotools make it very easy to do).
But you can't stop this discussion as long as people are hurt by your decisions.
Sure, I can't stop people from discussing anything they feel strongly about, but I would hope that such discussion would also be reasonably grounded in practicality. It's not practical to expect Jeremy to devote time and energy to maintaining these tools when he has so much else to do. It's not practical to expect that "simply shipping the Tiger bits" will be zero work since once they're shipped, they're expected to work, and Jeremy has already pointed out many times that the Darwin configuration files have bit-rotted and will only continue to get worse over time. It's particularly impractical to argue that Apple should maintain what X.org has already dropped - if you want to argue this with the folks ultimately responsible, why even do it here? There is a list called email@hidden where the folks ultimately responsible for this decision hang out, so if you want to argue this at the source, that's where to go.
Sometimes the most practical direction is forward, and that is the direction that Jeremy is trying to go in. If you want to fix anything in X11 on Leopard, please fix the stupid typo in /usr/X11/lib/libXrandr.la that breaks gnu libtool-based software that uses -lXrandr. (In case you forgot, the libXrandr.la file shipped with 10.5.2 lists a library name that does not exist in the X11 shipped with 10.5.2).
Jeremy has fixed a LOT of bugs in X11 and I'm sure this is on his TODO list, assuming that a bug report has been filed.
- Jordan
|