Quality Control in Ubuntu Server

[Patrick Goetz]

Last week I tried to upgrade a battered old Debian server running linux 
2.6.3 to a sparkling new machine running Ubuntu Karmic 64-bit AMD 
server.  The old Debian server was working perfectly, but had basically 
run out of disk space.  After 12 hours, I had to concede defeat, and 
rolled back to the Debian machine.  The issue is documented here:

https://bugs.launchpad.net/ubuntu/+source/samba/+bug/521085

Yes, believe it or not, environments running linux servers with Win XP 
clients like to be able to use their XP desktops even after the server 
has been upgraded.  This bug is a deal killer in this context.

The truly irritating thing about this bug is that it is identical to the 
one marked as "Fixed" here:
https://bugs.launchpad.net/ubuntu/jaunty/+source/samba/+bug/397699

Some other issues (an unpatched kernel oops which affects 
Hardy/Intrepid/Karmic but which is marked as Fixed in Lucid) has my 
colleagues muttering things like "This is why people use CentOS for 
servers -- it's stable!"

So, this is my thought:  maybe there should be a published checklist of 
specific core services (the kernel, LDAP, NFS, filesystems X-Y-Z, MySQL, 
Apache, Samba, Dovecot, Postfix) which are thoroughly tested *in* 
*specific* *contexts* before an Ubuntu server distro is released?  Back 
in the pre-2.6 days people would publish lists of hardware components 
known to work with particular linux kernels.  The idea is similar to 
this:  Publish a checklist of specific functionality:

    LDAP authentication in Samba
    SMTP TLS authentication using Dovecot SASL
    Saving/Reading 1TB files on Ext4
    etc.

which has been tested by someone before the server distro is released. 
Is this realistic?  Motivated users like myself could sign up to agree 
to test certain things, distributing what obviously is a rather 
prodigious task.  The published list would include configuration details 
helpful to someone trying to get some particular function working on a 
server.

Most of the time it only takes one or two occasions of serious wheel 
spinning (like I experienced last week) before someone throws up their 
hands and starts looking for an alternative solution.