I leave the project

Kevin P. Fleming kpfleming at linuxfromscratch.org
Wed Jul 7 13:28:01 PDT 2004

Matthew Burgess wrote:

> I take issue with that.  I don't think we should ever ship with a known
> bad configuration, at least not for the *majority* of users.  It is
> quite possible that the configuration we supply is not the correct or
> most useful for some users, and where such issues are known they should
> be documented and pointers provided to where further information
> regarding configuration options are detailed.  I've seen *nothing*,
> recently, regarding broken configurations, and there's no open bugzilla
> issues regarding it either.  If we aren't made aware of the issues, how
> are we supposed to address them?

And I take issue with this paragraph. I don't really want to participate 
in this conversation anymore, because the editors have decreed what they 
are (or are not) willing to listen to, but I have respond to this statement.

Alexander, Ian and I have been discussing the pitfalls of using unstable 
as it stands today with any significant number of modules in use. The 
archives will show that there are a number of issues that could (and 
will) trip up users who are not already experienced with them, which 
they are likely not to be since there are few (if any) distros using 
2.6/udev/hotplug and LFS-6.0 is likely to be their first exposure to 
this combination of packages.

In spite of being well aware of these issues, and the trouble they will 
cause for each and every LFS-6.0 builder who chooses to build most (or 
all) of their hardware drivers as modules, the "editors" have decided 
that the book has no place to either help these people with these issues 
(which was Alexander's choice) or to warn them of the issues so they can 
avoid them until they have time to learn how to deal with them (my 
choice). I have been told point-blank that if LFS builders are not 
already aware of these issues and how to deal with them ("module loading 
is a basic part of kernel knowledge" was what I was told, even though 
introducing udev and hotplug into the mix radically changes things), 
then they shouldn't be building LFS.

That decision means that LFS-6.0 is targeted at people who are already 
experienced with 2.6, udev, hotplug and the like, and how they interact. 
If that is the case, the "required prerequisites" for LFS-6.0 need to be 
radically altered, and will have to direct people to read non-existent 
documentation on how these packages can interact.

Saying that you are not aware of these issues when they have been 
discussed for the past month or longer on lfs-dev, blfs-dev and IRC is 
radically misstating the situation, IMHO. The editors have been made 
well aware of these issues, but they have chosen a path to deal with 
them which some of us do not agree with. In Alexander's case, since the 
bulk of his recent work was specifically around these issues, he chose 
to leave. Another poster mentioned "brain drain", and they are correct: 
it has happened, and it will continue to happen. Some of it is natural 
attrition, but I personally believe that most of it is not.

More information about the lfs-dev mailing list