[Gnu-arch-users] Re: a brief note on "delta compression"

[Stefan Monnier]

> I don't know how much the actual size of the _deltas_ would be reduced by
> e.g. summary deltas (as dicussed here before), but with tla it usually seems
> to be the number of changesets applied that's the killer in this case.

Indeed, the problems seems to be the time it takes to apply each changeset.
There is no doubt in my mind that this can be sped up by an order of
magnitude.  IIUC it is currently slow because it does a full inventory for
each and every changeset, i.e. it traverses the whole tree even if it's
only going to patch a handful of files, and even though it just did a full
inventory a second ago and it knows nobody else messed with the tree in the
mean time.


        Stefan