LFS and BLFS

Alexander E. Patrakov patrakov at gmail.com
Sat Dec 15 09:36:52 PST 2007


2007/12/15, Randy McMurchy <randy at linuxfromscratch.org>:

> I can't help but think that something like this would
> turn BLFS into just another distro. And an exceptionally
> difficult one at that.

I think I disagree with "_just_ another distro". It becomes a
"documented distro".

> But you don't understand Ag, the package instructions *do* work.

Well, maybe. But we don't know that they work, and that's more
important (but see below - you say that you rely on general public to
report breakage).

> We are only providing instructions to build the package. If it
> doesn't work after a successful build, we can only hope that a
> member of the community lets us know this. If we receive notice
> that a package does not *work* as designed, then we'll take an
> appropriate action.

IMHO, this is slightly inconsistent. What you are trying to say is
that we receive such notices via the mailing lists and incorporate
them to the book - i.e., we only act as men in the middle between the
book and the general public. As the LiveCD development process has
shown, this just doesn't work: the NVidia driver version 100.14.11 was
first added in r1937 without libwfb.so symlink (i.e., completely
broken, but with working alternatives in the form of earlier versions
that don't need the symlink). It took two months for this "completely
non-working package" bug to be reported, even though I know several
people who at that time attempted to build the CD after every SVN
commit.

But if this "men in the middle" model is really the case, and the
development is in fact supposed to be community-driven but not
completely self-controlled, then there is a better model: wiki with
"approved revisions", like http://en.citizendium.org/wiki/Main_Page.

-- 
Alexander E. Patrakov



More information about the blfs-dev mailing list