[tex-live] Stable vs. Unstable/Testing Update Repositories?

[Victor Ivrii]

2010/2/24 Zdenek Wagner <zdenek.wagner at gmail.com>:
> 2010/2/24 C.M. Connelly <cmc at math.hmc.edu>:
>>
>> The beamer issue is just an example of the problems that can
>> happen when you're dumping brand-new untested code into a update
>> stream that's easily accessed by naive users.
>>
> Wrong! As turned out, the geometry package was tested and bug free.

A bit too strong statement: let say "bugs were not found so far".

> It just turned on the beamer bug that was there unnoticed for a long
> time.

I am a bit uneasy with statements "people on deadlines should not
update". There are some people who rely upon some big server
administrated by sysadmin and at any given moment someone is on
deadline. This is exactly the reason why system administrators in our
dept maintain teTeX and allowed me to maintain TL. At least they know
that they "did not changed nothing" when one of my esteemed colleagues
screaming bloody murder requires explanations why some file which
worked for the last 200 years stopped working (in 99.999% cases they
made "minor" changes). So sysadmins want some assurance.

May be limited rollback feature would be helpful? For example keep
previous versions of some major packages?


Just thought

Victor



========================
Victor Ivrii, Professor, Department of Mathematics, University of Toronto
http://www.math.toronto.edu/ivrii