On Thu, Mar 6, 2014 at 1:35 AM, Allan McRae <allan@archlinux.org> wrote:
I'm assuming that is because that most of the time/memory is used applying the diff rather than decompressing it given it is quite small.
I know memory usage and speed were issues when it was considered adding -9 to package compression. My concern is this also applies to deltas. However, given the reconstruction of a delta is reasonably computationally intensive, I'm not sure this restriction applies here.
Anyone else care to comment?
I've been running repo-add -d with XDELTA="-9 -S djw" to tease out better ratios. That said, with the current system delta + XZ is kinda suboptimal since applying the delta requires recompression of the package and XZ's compression times are so awful.