On Mon, Oct 01, 2012 at 08:38:48PM -0500, Bruce Dubbs wrote:
> Ken Moffat wrote:
> 
> >   Release early, release often.  For major projects, weekly -rc
> > releases, and stable point releases as-necessary, is a good thing.
> 
> The release early, release often approach can be overdone.  How is a 
> typical user supposed to know when a change is significant?  I have no 
> problems with labeling a package as an intermediate point, but a 
> 'release' is normally assumed to be fully stable unless otherwise noted, 
> and once a week is just too much.  Why bother to use a version control 
> system?
> 
 Thinking particularly of the kernel, a lot can change in a week
during the -rc stage.  Mostly, for the better.  The fixes that also
apply to previous _releases_ also get into stable after they have
been in Linus' tree.

 For many packages, you can read the release notes - for the kernel,
there is usually something out there listing the major details, even
if the git log is a bit impenetrable.

 Anyway, if there is a bad commit it can be reverted or otherwise
fixed once discovered - that's what version control permits.  But I
disagree that any release is "fully stable" - usually, it just means
that nobody has found bugs that the developers/maintainer think are
important enough to delay a release.  All software has bugs, many
lie undiscovered for years.  In applications, many other bugs hang
around, sometimes for years, because the developers don't use linux.

> I've wondered, in the case of the kernel, whether is would be beneficial 
> to release the drivers as a separate tarball on a different release 
> schedule.  I think that's where most of the changes occur.  In an 
> uncompressed kernel tarball, the drivers are 260M of 532M total.
> 

 Drivers change for two reasons : fixes, and changes in the rest of
the kernel - the interfaces are not stable (I nearly wrote API, but
drivers aren't an application).  Either way, a few people use
Linus' git tree for testing - a few more more people test the -rc
versions when it looks as if they might be stable enough to reduce
the pain (e.g. -rc3 or -rc4).  Anyway, modern disks are big :)

> >   I don't follow systemd, no doubt some of the changes are important
> > for the project, but without monitoring it I can't guess which, if
> > any, impact the udev part.  I'm still hoping that standalone udev
> > will gain traction.
> 
> That would require a major attitude change from the developers.  IMO 
> they have a vested emotional interest in not separating udev.
> 
 I meant the fork created by gentoo users.  I'm encouraged to see
that they seem to be picking up the useful things from "upstream".

ĸen
-- 
das eine Mal als Tragödie, das andere Mal als Farce
-- 
http://linuxfromscratch.org/mailman/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to