Le mercredi, 4 février 2015 à 00:30, Christophe Troestler a écrit : > Regardless of the final scheme adopted for error handling, it seems > that a type such as Result.t is important to share among libraries. > Since depending on Core_kernel will stir debate, why not release an > OPAM package "result" declaring Result.t together with a few useful > functions (including monadic combinators) and ask Jane Street if they > would agree that Core_kernel reexports it in order to be compatible? > I know that one could dispense with that using polymorphic variants > but, apart from the better efficiency, I think the slightly stricter > type checking would also be beneficial. > > What do you think? I'm not sure that's very useful since often you will want to work with this type over a monadic concurrency library and/or provide combinators that interact with a logging library (e.g. see the assemblage combinators I linked to earlier). It seems to me there will be very little left and loose coupling between packages seems fine to me in that case. I rather think that mirage itself should define a library with all the good tools to enforce its chosen error handling scheme.
Best, Daniel _______________________________________________ MirageOS-devel mailing list [email protected] http://lists.xenproject.org/cgi-bin/mailman/listinfo/mirageos-devel
