This looks like a very interesting project, but I am wondering what the goal is, is this supposed to be compatible with a large fraction of packages on Hackage? Or is that simply not a goal?

For example, parsec-2.1.0.1 has tabs and carriage return in its cabal file. I was able to adjust the parser to accept that. But it also uses braces, which I hadn't seen before. And now I don't know whether it makes sense to start to have MicroCabal parse this.

Since I suspect that the code will be more minimal if unnecessary features are avoided, it's a trade-off: Either the code is uglier and more compatible, or it is prettier and less compatible. To align contributors to the project goals, it would make sense to specify the goals in the readme file.

MicroCabal commit 85cbd51 fails on e.g. https://hackage.haskell.org/package/penny-0.32.0.10/penny.cabal with an error about the tested-with line.

For example on https://hackage.haskell.org/package/epic-0.9.3.3/epic.cabal, MicroCabal reports an error on line 21:

found: 'e' expected: ','

Tested with commit 85cbd51

$ mcabal install simple-cmd
Building in /var/home/petersen/.mcabal/packages/simple-cmd-0.2.7
Building library simple-cmd
mcabal: /var/home/petersen/.mcabal/mhs-0.9.11.0/packages: getDirectoryContents:openDirStream: does not exist (No such file or directory)

(also this is actually 0.9.13.0 : - )