 |
Stephan, do I remember correctly, that you've taken a look at this? Did it look promising? |
|
|
The problem with xdelta is, as far as I know, that it needs the full old data beside the new (changed) data to compute the delta. That conflicts with the goal to minimize network data transfer. It would be necesary to transfer the full new data via network to where the old data resides enable to compute the delta using xdelta. It looks like only rdiff (librsync) supports computing the delta using only the much smaller signature of the old data. On the other hand, the deltas generated by xdelta are a lot smaller than rdiff deltas. For a more detailed comparison: There's a master thesis about "Binary Differencing for Media Files" from Vladimir Komsiyski at http://dare.uva.nl/document/490827 So when considering minimizing network data transfer more important than minimizing delta size, rdiff (librsync) would be the better choice. |
|
|
We try to tackle this via either GSOC or some thesis work. |
- Anonymous
