My suggestion takes care of all string matching collisions:
Let's say you have
-rw-r--r-- 1 hacksaw root 446031 Apr 5 1996 zephyr
and
-rw-r--r-- 1 alanc wheel 446031 Apr 5 1996 zephyr
In the same directory.
Let's then assume some random guru can see both, say Linus.
Linus would see:
-rw-r--r-- 1 hacksaw root 446031 Apr 5 1996 zephyr
-rw-r--r-- 1 alanc wheel 458991 Apr 7 1996 zephyr
In a long listing, indicating he had two files that were named the same
thing but owned by different folks. If he wanted to refer to on of them,
he might say:
more zephyr:alanc
Or something equivalent that doesn't screw up shells. Their real stored
name would be zephyr, but namespaces would have scope by owner. In fact
you could make it zephyr::alanc, just for the mnemonic to c++
Admittedly, this is not an easy change. Filename buffers would have to
be carefully lengthened, and extra checking done to make sure the name
didn't grow too long for the filesystem.
Of course, the algorithm would need to treat the given filename as a
regular, unqualified filename first, in case someone has named a file
foo::bar, in which case the FQFN might be foo::bar::baz.
Have I muddled your head sufficiently?
-- Hacksaw- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu