On Sun, Oct 23, 2016 at 11:37 AM, Allan McRae <[email protected]> wrote: > What can we conclude from this? > > 1) no operation reads a single "desc" file > 2) -Qo is the only operation to gain from efficient reading of all > "files" files. (-Ql may also gain, but is probably output bound) > 3) having "desc" and "files" files separate and only loading these files > if we need it is a gain. > 4) all operations on mtree files require lots of disk access, so there > is not much gain in increasing the efficiency of reading these.
It's pretty intuitive too, considering "desc" files cross reference other packages, while "files" is private to a package. It makes sense to group or link all "dest" files into a web, while "files" files are mostly isolated.
