Irfan Alibay

1) pulling from a deployed version is a bad idea - deps change and so do pins, it might work today but it's a very flaky solution at best. 2)...

@RMeli the issues with pyproject.toml for develop builds is that this builds a purely "minimal deps" build - i.e. any conda-forge only dependencies won't work (e.g. openmm or rkdit that...

I'm sorry but this thread doesn't really make a lot of sense to me right now - we seem to be looking to create fancy solutions to a problem I'm...

I did the above in #371

Can we maybe change the warning to say "if you're using MPS, you're probably fine, if not, then consider it"?

I vote for 3.

Just wanted to weigh in here, especially given the proposed patch from: https://github.com/openforcefield/openff-interchange/issues/1335#issuecomment-3328738234 > I didn't know that OpenMM needed to be told these forces are periodic. For normal bonded...

Looking through historical issues, Peter Eastman highlights two things: 1. Performance issues 2. The bond calculations being incorrect if they shouldn't be periodic. Probably a question for the OpenMM issue...

Thanks so much for including me, I agree to being an author. I'll give things a review as soon as possible.