4 ms·

Given the target audience of Anaconda, I can see how BSD compatibility doesn't matter to them. I guess it would take some major lifting because much of what the

Given the target audience of Anaconda, I can see how BSD compatibility doesn't matter to them. I guess it would take some major lifting because much of what they do is wrangling the compiling environment. Supporting Windows is hard enough. That's not "being toxic", you just can't be everything to all people.

inferiorhuman 4y ago
Nah it's toxic. It's not just that Anaconda won't provide binaries, it's that you can't even build a project that uses Anaconda on not-Linux/Windows. In terms of why you shouldn't tell someone to simply use Anaconda, that's pretty high up there. As tedious as javascript package management can be, python has consistently given me more trouble.
- bayesian_horse 4y ago
  I think that's an "Am I the Asshole" situation, and you are getting it wrong. BSD is niche, especially in scientific computing. Saying anaconda is worthless because its maintainers don't expend a ton of effort on making your particular niche more convenient is the actual toxicity...
  - inferiorhuman 4y ago
    Saying anaconda is worthless because its maintainers don't expend a ton of effort on making your particular niche more convenient is the actual toxicity... Good thing that's not what I'm saying. The problem with Anaconda isn't that it doesn't support BSD (or whatever), the problem is that by using Anaconda you prevent a project from building on anything that Anaconda doesn't support. It's a poor design that simply saddles python with vendor lock-in and, yes, that is toxic. It's a ridiculous non-solution given that even in the data science context most stuff is POSIX compatible. If I can get R and Python on their own to work without hassle, there's no reason a mere package manager should throw up unnecessary walls.
    - bayesian_horse 4y ago
      I think this discussion is pointless. No "project" depends on conda/anaconda, because a package manager, in Python, is a means of deployment at most. If indeed you have some project that comes with a conda environment description (or dockerfile or whatever) then all you need to do to make it run on BSD is figure out a way that works on BSD. Of which there are certainly many. It's really just that conda environments are pretty damn convenient for Windows and Linux in the context of numerical libraries.

inferiorhuman 4y ago

Nah it's toxic. It's not just that Anaconda won't provide binaries, it's that you can't even build a project that uses Anaconda on not-Linux/Windows. In terms of why you shouldn't tell someone to simply use Anaconda, that's pretty high up there. As tedious as javascript package management can be, python has consistently given me more trouble.

bayesian_horse 4y ago
I think that's an "Am I the Asshole" situation, and you are getting it wrong. BSD is niche, especially in scientific computing. Saying anaconda is worthless because its maintainers don't expend a ton of effort on making your particular niche more convenient is the actual toxicity...
- inferiorhuman 4y ago
  Saying anaconda is worthless because its maintainers don't expend a ton of effort on making your particular niche more convenient is the actual toxicity... Good thing that's not what I'm saying. The problem with Anaconda isn't that it doesn't support BSD (or whatever), the problem is that by using Anaconda you prevent a project from building on anything that Anaconda doesn't support. It's a poor design that simply saddles python with vendor lock-in and, yes, that is toxic. It's a ridiculous non-solution given that even in the data science context most stuff is POSIX compatible. If I can get R and Python on their own to work without hassle, there's no reason a mere package manager should throw up unnecessary walls.
  - bayesian_horse 4y ago
    I think this discussion is pointless. No "project" depends on conda/anaconda, because a package manager, in Python, is a means of deployment at most. If indeed you have some project that comes with a conda environment description (or dockerfile or whatever) then all you need to do to make it run on BSD is figure out a way that works on BSD. Of which there are certainly many. It's really just that conda environments are pretty damn convenient for Windows and Linux in the context of numerical libraries.