<mbakke>zimoun: yes! I used it initially, but ironically it hampers reproducibility because channels are embedded, and there are many channel/commit combinations that can lead to the same output ... so a would-be reproducible build isn't because of slightly different channel.
<mbakke>there is a note about that in the manual :)
<zimoun>mbakke, about the note, yes, but I have never hit a real-world example. :-) Thanks for the explanations. Hum, I am still missing a point, since the issue should be fixed by the time-machine channels.scm.
<mbakke>zimoun: channels.scm refer to the master branch, but the provenance saves the exact commit ID
<zimoun>ah indeed, it is not the output of “guix describe -f channels”