locales, nls - supportable or not?

Alexander E. Patrakov patrakov at ums.usu.ru
Sat Nov 18 22:06:42 PST 2006

Dan Nicholson wrote:
> No one here understands glibc. Should that be removed from the book?

No, because the current level of understanding (explanation of 
./configure switches, list of installed files and trivial configuration 
instructions) is complete enough for LFS purposes. Almost nobody asks 
questions other than build problems caused by deviations from the book, 
and almost all questions are answered (the exception is about SCSI 
headers - but it would require a programmer to answer this). The problem 
doesn't exist here, IMHO, because the configuration instructions are 
trivial and the same for all users. The initial nsswitch.conf can be 
copied and pasted, and those who need to deviate (i.e., keep users in 
NIS or LDAP) can find the needed information in Google. It is also 
important that Google doesn't list outdated and/or wrong documentation 
in this case.

> The new kernel headers
> are causing breakage here and there. Does that mean that one of the
> editors needs to have C and kernel development experience?

C - yes, kernel - optional for this problem (because it's userspace what 

> To me, all I can hope for is that it works well in my use case
> and if it doesn't, hopefully the issues can be tracked down.

This is a valid viewpoint. However, for UTF-8, the user could not track 
down his issue.

> Look at BLFS. Is there any possible way that between Bruce, Randy, DJ,
> Ken and I that we have an in depth understanding about how
> Kerberos/LDAP/SSL/SMTP/X11/etc. work? Not a chance in the world.

You underestimate them. Randy is the author of the heimdal.txt hint 
(which not only lists instructions, but also the steps needed to verify 
the user's setup), Jim has a sophisticated Courier MTA setup (that was 
removed from the book exactly because nobody else could maintain it).

> What makes LFS fun for me is having the chance to build things myself
> and investigate them with a group of like minded people. I'm not here
> for high level support. If I have an issue with my mail server, I put
> it on blfs-support and hope for the best. Maybe Ken or someone else
> has seen the same thing.

Indeed, the support lists contain valuable information, as they should.

> For this particular case, maybe you're right. Maybe this is too far
> removed from the current set of editors skills to get right with all
> the patches and use cases. If people would rather not have UTF-8 in
> the book because it's not working and they can't get decent answers
I read "not working" as "not working (even because of their own 
misconfigurations and deviations from the book, as it was the case)"
> from the editors, then I would be for removing it.

So let me summarize the above. You say that for topics like 
Kerberos/SMTP/X11, you are satisfied with the level of support on the 
lists, because there are people skilled in these areas there. But for 
locale support, there are not enough such people, and sending people to 
distros (away from LFS) would be the best advice.

> What I have an issue with here is that the level of support you expect
> is not in any way realistic.

The only issue is that, before I introduced UTF-8 and before Greg 
Schafer left the project, LFS and BLFS actually had this level of support.

Alexander E. Patrakov

More information about the lfs-dev mailing list