nanos
nanos copied to clipboard
Hook up path canonicalization and add a namecache
We should hook up the path cannibalization code I wrote to open(2). At the same time it would be nice to implement a trivial namecache that translates long path names to tuples so we don't have to traverse each time.
How much does the traversal cost? I guess not much. rtree is already a cache, so it is a question of avoiding the traversal. I would not do this unless this turns out to be a performance bottleneck.