[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Fedora Life Cycle



Greg DeKoenigsberg escribió:
On Mon, 22 Oct 2007, Gian Paolo Mureddu wrote:

IMHO, it's far more interesting -- and useful -- to make upgrades work flawlessly.

I couldn't agree more with you on this! Theoretically upgrades shouldn't need to be too difficult, heck you can sort of do them "by hand" if you know what files you need and more specifically, what /parts/ of the files are needed... I'm specifically talking about passwd, shadow, group & gshadow, and paths such as /home, /root, etc. Of course there's also the "individual applications' config files, which can still be worked out. I've been thinking about this and it shouldn't be too difficult, but have been told time and time again that such a feat is impractical and nonsensical in the long run. I'm not convinced, but, then again it could be made possible for an automatic upgrade process to also be clean enough... I'll give it a bit more thought and maybe post an RFE on Bgzilla about the issue.

The current thinking (which I agree with) favors Anaconda-based upgrading with a new and improved online component, rather than yum-based upgrading. Take a look at the following feature proposal:

http://fedoraproject.org/wiki/Features/PreUpgrade

Think about your use cases in the context of this feature, and maybe add any comments you have to the wiki page itself.

--g


Indeed, what I have thought of is /primarily/ for Anaconda. Yum could also be involved, but first one thing at a time, a better finer control over the upgrade from within Anaconda for tasks where GUI does make sense for this would be a priority before making "networked upgrades" work with Yum. Simple stuff like being able to "choose" which users you want to migrate, nuke everything in the disk, except from chosen directories (/home/{user1,user2...n}, /root, /usr/local, etc), allowing the distribution to first get in place all the files and configurations it needs, THEN modify whatever files with only the information "salvaged" from the previous one. Getting that to be as clean and painless as possible should be a priority for Anaconda upgrading. Will take a look at the Wiki and see what other ideas can I toss around.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]