FC4 good new tech, bad legacy support

Matthew Saltzman mjs at ces.clemson.edu
Fri Jul 1 19:42:54 UTC 2005


On Thu, 30 Jun 2005, Les Mikesell wrote:

> On Thu, 2005-06-30 at 15:27, Richard Kelsch wrote:
>> I still stand by my claim that FC4 fails
>> the intentions of the project.  Nevertheless, I know it will be fixed
>> eventually, perhaps FC5.
>
> But you could bet that it would not be fixed if it wasn't released in
> its current state so people could fix it.   That's the point of the
> fedora releases - it is supposed to meet the usability intentions
> by the *end* of a release, when the effort shifts to a new batch
> of code and the updates to this one stop.  Since there are 3 prior
> releases you can get a pretty good idea how this works by looking
> back at the updates that made the other versions usable.

I've been using Linux since Red Hat Linux 3.0.3, and I can remember 
threads with exactly this theme and the same points of view represented 
for every major component upgrade including:

- a.out-format object files to ELF object files.
- kernel 1.something to kernel 2.0, to kernel 2.2, to kernel 2.4, to
   kernel 2.6
- gcc 2.95 to gcc 2.96-redhat-special to gcc 3.x
- libc5 to glibc
- linux threads to pthreads

and now

- gcc 3.x to gcc 4.0

Every time, Red Hat Linux--and now Fedora--was out front in moving to the 
new technology.  Every time, many things broke in early releases.  Every 
time, people bitched and moaned and predicted the end of Red Hat, the end 
of Linux, or the end of the world.  Every time, the broken stuff got fixed 
within a few months at most.  And the Linux world is better off in the 
long run for every one of these changes.

It would be nice to be able to make changes incrementally, but it's not 
always possible.  Kernels, compilers, and libraries affect almost every 
critical component in the system.  When you convert, you have to convert 
everything.  That's why these changes are made on major release 
boundaries.  (In the old days, RHL had major and minor releases to mark 
just such extensive infrastructure changes.)  When these kinds of changes 
occur, those of us willing to endure some "leading-edge windburn" will 
live with the issues and get them fixed.  Those who can't afford to, can 
chalk the early versions of these releases up to a "failure of robustness" 
if they like, as long as they realize that this is how progress gets made.

Richard's problem is likely that he's trying to build a Perl module that 
needs to integrate with a Perl built with gcc4.  He's well and truly stuck 
because the module won't build with gcc4 and it won't load if it isn't 
built with gcc4.  My recommendation for Richard is to sit out FC4, at 
least for now.  When the Perl module he needs gets updated so that it 
builds with gcc4 (and it will, perhaps even soon), he can migrate if he 
chooses.

(Richard's other problem is that I can never follow the attributions in 
his e-mails, because quoted material isn't clearly indented or set off 
with any visual marker like the '>' you see in most posts, but that's 
another story...)

-- 
 		Matthew Saltzman

Clemson University Math Sciences
mjs AT clemson DOT edu
http://www.math.clemson.edu/~mjs




More information about the fedora-list mailing list