regarding the TH description. I'm missing the following items, which would help understand this better (I think):
short introduction what iserv does and why asterius can't use it (highlight the specific issues iserv poses to asterius)
It seems like you've added a bunch of hooks to ghc so that you an intercept the ghc <-> iserv communication. Why could those hooks not be in a middleware process? ghc <-> middleware <-> iserv? Wouldn't that get all the messages as well and could do the linking? I'm sure I'm missing something integral here, but it's not clear from the documentation what I am missing. It just reads (to me) like we are shoving lots of logic into ghc (ahc to be precise) that seems like it could live in that middleware process as well and keep the existing iserv api in tact and reduce the changes necessary to ghc?
ghc
fork, which are mainly related to hookediserv
invocationsnode
session management and evaluation: how do we send compiled splices tonode
, trigger evaluation and fetch results