For 5 years, the ROracle #rstats package provided by #Oracle to CRAN has been broken on Windows. Some nice fellow has finally caved and posted a fork that actually works. Good for him! But it exposes an underlying problem,
I was going to post a snarky comment about Oracle scoring free labour off community desperation, but the SO comments suggest the maintainers tried to upload a fix to #rstats CRAN and failed, and have since given up on CRAN entirely - new working versions of ROracle are on the Oracle website.
Baffling that the package fix was apparently rejected but that the CRAN version wasn't archived, leaving it in limbo and giving an unfair impression that the Oracle maintainers don't care about their work.
@obrl_soil
As much as I don't like Oracle (mostly because of what they have done in the past), I havet o agree that CRAN's review process suffers from single-person-blockage. All your submissions come down to decision of one person. The #rOpenSci process is more democratic and way more transparent, but it is excruciatingly lengthy. I personally like a middle ground approach (transparent communication, more than one moderator for package submission, quick responses)
@Mehrad I don't have serious beef with that tbh, the gatekeepers aren't usually flat wrong. the thing I find odd is how the broken version has been allowed to remain, given that most of CRANs policies and general vibe over the past 2-3 years seem aimed at trying to keep their exploding package volume under some kind of control. It's another symptom that their operating model is just overwhelmed imo
@obrl_soil@Mehrad As long as it builds successfully on all of the tested platforms it will not be archived. If it starts failing the maintainers will be told and given a timeframe to fix it.
I also think that as long as an accepted package build and works and is not flagged for malicious code, it should stay available.
Regarding the CRAN review process, I think transparency and public review is the key and is the main missing part in the current system. Imho f a package is submitted to be public, then the admission and review process should be public too (same way that scientific publications' review process should be transparent and public).
@elinwaring@Mehrad it builds but doesn't install, which should make it a candidate for removal just as a build failure would. Especially given that the reasons for the installation failure are well known coding errors, not some esoteric OS issue.
@obrl_soil
That's interesting. I've never seen a package that builds correctly but did not install. I know that for my own packages on CRAN, the windows build it done by another person and it takes about a day or two after I submit the new version to CRAN. I don't use Windows though, so I have no clue on how complex packages internally work in that environment.
I suggest emailing Uwe Ligges who is maintaining windows binaries for packages:
Add comment