Hello, We need to change something. I am just not exactly sure what this is. Since software/workflows that we cover in Debian has increased in complexity, we have come up with salsa. And we are increasingly automating our testing. The package review has not yet seen any methodological change but we expect it to just somehow keep up. And complex software having some essential module waiting in new is - unfortunate for everyone. Just to give you an example, I am packaging for the same biological workflow since two or three years now. What keeps me going? Read through https://bcbio-nextgen.readthedocs.io/en/latest/ and ask yourself if you think that should work on Debian - it works via conda on Debian, but that is not what I meant. Why is this difficult? For many reasons. One is that it uses software that is written in "nim", which is not so very much known, yet. But this is also how Debian benefits from this effort as a whole. A serial dependency on five or so non-biological modules it has. And the first (nim-unicodeplus) sits silently in the New Queue since 8 months.
Folks at conda do everything on github, including a peer review. For most scientific packages I sense this to be just fine. Maybe we could somehow stage our developments? A peer-reviewed (as in "get at least three signatures" maybe?) instant upload into a "periphery" distribution with a transition into a "as time permits" FTPmaster-scrutinized "main" distribution? Also, I find it somehow sad that FTPmasters after their ingenious isolation of a problem that would then be fixed real easy don't have the chance to just do the fix in salsa and have the package accepted from there. The workflow now is that the package goes back to the uploading developer who then fixes the typically trivial bit and the package impose the same work on the FTPmasters again, often with a llooonngg delay again. Instant trivial fixes would dramatically reduce the rejection rate, I presume, and may also increase the fun to be an FTPmaster. Best, Steffen
