bw123 wrote:
It doesn't matter if the more recent pkg is available if it doesn't work. You have access to zillions of pkgs from all over the linux world, but they don't all work on stable. It's an old argument you're making, "I mixed repos before and it worked, now it doesn't so something is wrong."
you were answering to this statement: "1. In testing, one has access to more packages and more versions."
No! I persist and sign. I have used many packages from testing. And the cases causing corruption are rare and I can be prepared for corruption using VMs or docker.
Installing ruby-all and ruby-all-dev on testing works very well. And you get more choice of versions than what the stable branch offers.
Concerning the other methods, the backports seem to have much fewer packages than testing. Besides, backports seem to have more dependencies to each other.
Actually, many people would do a better usage of installing the entire debian testing instead of debian stable.