@simonpj that all sounds good to me. I agree the paper should mention these things. Not only would I want to taking that citation of “architecture of open source programs” on the origin of dyn flags and expanded it as you say, I think it would be good to move it up and make a key part of the first history/background half.
Following up on our discussion with @simonpj on modularity “behind” per-component configuration records, I opened Draft: Separate C-- building in general from STG -> C-- in particulare (!8237) · Merge requests · Glasgow Haskell Compiler / GHC · GitLab which internal to a single component (a C-- “builder” monad, used by each of the the C-- parser and the STG → C-- pass) uses Has* classes to do more fine-grained modularity without either a profusion of “subset” records or a profusion of (positional) arguments.
I am curious what anyone has to say about such a second step beyond just breaking up DynFlags!
1 Like