3 ms·
He missed some: First try a zlib of the entire thing, not each version, since it will effectively find all the common lines and not repeat them. Next, instead
by ars 18y ago
He missed some:
First try a zlib of the entire thing, not each version, since it will effectively find all the common lines and not repeat them.
Next, instead of pure keyframes, let the keyframes themself be chained. That should save a ton of space, without being really slow. (And then recurse if you like - every ten keyframes, add a supra-keyframe, etc. recurse all the way down.)
- jrockway 18y agoYeah, the zlib of the entire thing would be interesting. Also interesting would be gzip(orignal + set of deltas). I would also like to see how his data did with a git packfile, and when stored in a svn repository. (This would answer an important question: Are the theoretical ideas better than what is currently used in practice?)