On Sun, 9 Nov 2008 13:17:31 -0800 "David Fox" <[EMAIL PROTECTED]> wrote:
> I was helping to update a friends (old) p3 box that ran an old version > of etch, needed a bunch of security updates and kde and some other > stuff installed last night. I got all the updates installed except for > two things - open office (not sure he'll be using it, he wanted squid > set up primarily) and there were some errors about not being able to > access or write some font directories. Yeah, things like that can happen --- not very important maybe, but if you happen not to be able to install or update exim or cyrus on the mailserver over the weekend, it's not so nice when people come back to work on Monday and find out that they can't read or send their emails. And I'm talking about upgrading Sarge, involving switching from exim3 to exim4 as well as switching to a newer version of cyrus that required to convert all the mails to a differently organized format. Updating the webserver from Potatoe to testing-after-Sarge some time after there were no more security updates for Potatoe is also something you don't exactly want to do (I didn't). Updating an early amd64 distribution on the file server that was, at the time of installing, hosted on Alioth, to the later amd64 that was an official release is pretty much impossible. So the point is that upgrading from one release to another one is a leap that can not be undertaken lightly, if at all. That a stable release does not break doesn't mean that upgrading from one stable release to another doesn't break it. That is a problem you don't have when you run testing and keep it updated. You may have other problems running testing (I didn't), but if things really go wrong, you can still update to unstable from there. What is best to use still depends on the requirements, of course. If you really need it rock solid, run stable and make plans how to perform the upgrade to the next stable release. Is there a way back, like from testing to stable or from unstable to testing? Anyway, I do have a lot of trust in Debian (maybe too much?). Attributing stable as "rock solid" is quite an understatement, considering that testing uses to be rock solid. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
