This month is going to be a bit more random than usual where git-annex development is concerned.

  • On Saturday, the Seven Day Roguelike competition begins, and I will be spending a week building a game in haskell, to the exclusion of almost all other work.
  • On March 18th, I'll be at the Boston Haskell User's group. (Attending, not presenting.)
  • March 19-20, I'll be at Dartmouth visiting with the DataLad developers and learning more about what it needs from git-annex.
  • March 21-22, I'll be at the FSF's LibrePlanet conference at MIT.

Got started on the randomness today with this design proposal for using git-annex to back up the entire Internet Archive. This is something the Archive Team is considering taking on, and I had several hours driving and hiking to think about it and came up with a workable design. (Assuming large enough crowd of volunteers.)

Don't know if it will happen, but it was a useful thought problem to see how git-annex works, and doesn't work in this unusual use case.

One interesting thing to come out of that is that git-annex fsck does not currently make any record of successful fscks. In a very large distributed system, it can be useful to have successful fscks of an object's content recorded, by updating the timestamp in the location log to say "this repository still had the content at this time".