David L. Mobley

I strongly agree that we want diagnostic regression tests, such as those that would have caught this. Certainly sudden changes in charges of 0.01 or 0.1 e ought to result...

Perhaps the 1-4 scale factor issue should be split into a separate issue, as this'll be a harder fix i think. Maybe this issue should just involve documenting it/mentioning it...

> Do we want SMIRNOFF to describe implicit solvents as well as explicit solvents? Yes > If so, should they be separate force fields, or should one force field describe...

Sorry, the "opt angle is way too small"? Which angle?

Ah, the "optimized angle", after geometry optimization.

OK, so for the record (cross-posting from Slack) this is a phosphonic acid with a proton and a sulfonic acid with a proton, and the proton ends up way too...

Right, that was the other thing I was forgetting: We had crashes with GAFF, too, when simulating sulfonic acids in such environments. I had thought we'd not have crashes with...

@jchodera see [SMIRNOFF paper SI](https://pubs.acs.org/doi/suppl/10.1021/acs.jctc.8b00640/suppl_file/ct8b00640_si_001.pdf) section 3 . These are small because AMBER formerly had them as zero, and for `smirnoff99Frosst` we essentially made them just nonzero enough to remove...

I'm in favor of this if we can easily make a generalized `LibraryChargeHandler` but I'm not sure we have that functionality. On the other hand it may be trivial by...

@mattwthompson you mentioned this on the list of things requiring feedback; what feedback is needed on this one?