Proofmode with Coq binders, and without strings
@janno mentioned to me some ideas for how one could make a proofmode that uses Coq names and syntax instead of strings. Just imagine, we could have nested Coq- and Iris-level intro patterns... that would cut many proofs in lambdaRust significantly. We could also not use strings, which may even help with speed. (Well, strings are not very efficient in Coq -- but I kind of doubt that's out bottleneck currently.)
I'll leave it to Janno to explain the idea, because I would totally screw that up.
I any case, even if we want to go for this, this is a long-term thing... definitely post-3.0, and definitely breaking everybody using Iris horribly (if we don't just keep the string-based stuff around).