Re: Starting a grad project that may change kernel VFS. Early research Re: Starting a grad project that may change kernel VFS. Early research

[Theodore Tso <tytso@mit.edu>]

Something which I urge you to think about is whether optimizing du -s
is really worth it; this is not at all obvious.  If it's going to cost
performance; if it's going to require non-backwards compatible changes
to filesystems; if it means introducing changes to the semantics of
hard links --- please *seriously* consider whether or not it's worth
it.

In what workload or use case is "du -s" something that gets done
frequently?  And is it really worth adding code complexity and slowing
down much more common operations, such as writing files?

If the goal is do a project that gets you a master's degree or a
Ph.D., ok, fine; there are plenty of academic degrees and honors which
are given for ideas that are completely impractical and/or useless in
the real world.

If your goal is to create a feature that will be accepted into
mainline, then you need to provide a lot more justification that this
is actually a good and worthwhile thing to do, and that the benefits
outweigh the costs (in code complexity, long-term maintenance,
performance regressions, etc.)

						- Ted