pylkko wrote:Actually, it is worse than that. Because testing packages are often older than backports. For example, if the argument is supposedly based on needing support for newer hardware, then probably we can assume that a newer kernel would help. However, the kernel in testing is now 4.9.6, while the kernel in backports is at 4.9.13. This is not the only package, of course, almost every single new package goes to backports before it goes to testing.
This is not true. Backports are, factually and by policy, taken from Testing. Only security updates may be taken directly from Unstable, which is also true for Stable.
If one looks at the differences between the versions in Jessie-Backports and Stretch (1): Today, 850 out of the 1803 available backports are at the same version in Jessie-Backports and Stretch, 633 are outdated in J-B, 252 are older in J-B, and only 7 packages in J-B are at a newer version than in Stretch. Among these seven packages is new kernel (the others are either blocked by the current freeze or have been removed from Stretch.)
And when one checks the kernel changelog (2), one notices that version 4.9.13-1~bpo8+1
was uploaded to Jessie-Backports eight hours
after its availability in Unstable. One has to imagine that such a fast upload was allowed only because of a security fix, even though there was no security announcement related to the kernel in the corresponding Backports Security Announcements mailing list (3).
In any case, if one picks any other backport, there is an extremely high chance that it was uploaded to Backports about one week after it had already been uploaded to Testing.
2. http://metadata.ftp-master.debian.org/c ... _changelog