Going back a bit:
AdrianTM wrote:Have you considered that maybe Debian Stable is not for you?
Actually I'm pretty sure that it
is for me. I evaluated all the major free distros and Debian Stable is unique. 1) I love the low bug count and the "tried and true" nature of the distro as a whole. This makes a huge difference to me as a
desktop user. In fact it perplexes me that no one other than sysadmins seem to appreciate the value of reliability. The fact that it changes less is also a bonus in that it reduces the amount of maintenance. My desktop is more of a workstation; I need it to get things done and every bug is a battle I don't need. I like the -stable philosophy, I think it's almost perfect. 2) Also it can be upgraded to new versions without forcing a re-install.
The stated downside to -stable is the obsolescence. My perception is that this is a necessary compromise, i.e. because system testing takes a lot of time - but this isn't necessary for ALL packages.
The backports concept makes a lot of sense to me now. I can upgrade from OO.o 2.0 to 2.4 without upgrading anything else. That is a powerful advantage, so far as it's possible, as it allows me to pick and choose which applications I am willing to sacrifice to get features. Perhaps this is even better than my original idea taken from FreeBSD, since backports is designed to upgrade on a case-by-case basis, and it is capable of reverting to -stable versions, rather than update the entire application base, which probably would lead to an unmanageable bug-tracking scenario.
However, Backports.org could stand to be improved substantially, IMHO:
1) There is a difference between making a command-line feature versus a GUI feature. I.e. backports is not presently an option for the Unix-disabled masses. For example, when I asked APT to upgrade openoffice-base, it initially guessed that I wanted to break openoffice-calc (and all others), so I had to guess and check to get my command right. Only a competent admin can use it like this.
2) Of the 3 items on my short-list for improvement, only 1 (OO.o) was actually carried in Backports. My long-list had about the same success ratio. Library problems may prevent backporting some packages, but I'm pretty sure Ardour doesn't fall in that category... so... stuff is missing?
I realize there may be practical limitations to supporting another repo - so perhaps it would make sense to limit bug response to broken builds. I'm willing to accept a Backport that comes with no warrantee but leaves me able to "downgrade" to the -stable version. This approach especially makes sense to me where the upgrade list is itemized.