For more information, see http://blog.websages.com/2010/12/29/recollections/

!!!!!!!!!!!!! DANGER !!!!!!!!!!!!!!!!!!!!!

This code is just a proof-of-concept. It needs a lot of work, and

probably converted to a dynamic language and not to suck in general.

This all must be one filesystem (or a CAS-backed apache tree), as we use hard links to de-duplicate

${somewhere}

• control <- [ control builds and maintains the repositories ] | + apache | + bin <- binaries for processing (ingredients -> preparation -> product) | + etc | + hooks | - README |
• ingredients <- [ ingredients are un-altered upstream sources ] | + cpan <- unpackaged perl modules | + cvs <- cvs checkouts from external sources | + git <- cloned git sources | + svn <- svn checkouts from external sources | + isos <- iso files not on mirrors | + mirrors <- timestamped mirrors of external package sources | | + centos | | + 4B903324 | | + 4BCC8AC2 | | + ... | | + 4CFD0510 | | + latest -> 4CFD0510 | | + cygwin | | + dag-rpmforge | | + debian | | + debian-multimedia | | + epel | | + fedora | | + debian | | + unattended | + misc <- items not from source control, tarballs, cpan | + source <- source (tar.gz) not packaged | + srpms <- source rpms not from mirrors |
• preparation <- [ preparation is where we turn our sources into repositories ] | + build <- space to create packages from ingredients | + recollection <- snapshots of working that have been repo closed and hashed | + working <- cloned from ingredients/mirrors, users can http PUT new content |
• product <- [ product is where we store the "frozen" repositories
  • data_cas <- recollection data permanently stored
  • data_idx <- recollection indexes permanently stored