Dimitry Golubovsky
If one tries to search Google:
it returns URLs of all Google-indexed .cabal files (currently 7- just this few?)
That's certainly not all of them. Just in the fptools tree we have: Cabal.cabal GLUT.cabal HGL.cabal HUnit.cabal HaXml.cabal OpenAL.cabal OpenGL.cabal QuickCheck.cabal Win32.cabal X11.cabal arrows.cabal base.cabal fgl.cabal src.cabal haskell98.cabal monadLib.cabal mtl.cabal network.cabal parsec.cabal unix.cabal
So if one makes such a request from a program, then parses the resulting HTML extracting <a> tags containing links ending with ".cabal", it would result in a directory of .cabal files locations suitable for furter processing. Similarly, if all locations of a particular package need to be found, package name needs to be added to the query string.
Using Google API might be an alternative, but it returns only up to 10 results (per their docs) which might be enough to find all mirrors of a single package, but not the whole list of available packages.
Publishing in such a directory might be simple: just place one's .cabal file in a place where Google can index it.
One weakness: too easy to spam.
We could possibly overcome that by using the cryptographic signing process similar to what I implemented for apt-secure in Debian. I've already talked to Lemmih a bit about this and he's working on it. Not sure if cabal-get really wants to troll google, though :) peace, isaac