>> Wouldn't it be possible to pass the filename information to userland
>> without loss and do the conversion there [...]?

> Note it would be much less of bloat to do it in kernel than doing it
> in userland - in kernel, the recoding structures can be shared via
> pointers; in userland, each process would have to read it into its
> own address space.

One person's bloat isn't necessarily another's.

I sure don't want to have to pay the penalty for permanently resident
large tables just to support the one or two occasions when I want to do
something with such-and-such a filesystem; I'd much rather pay a larger
penalty in userland, when it's needed, than pay the smaller (but still
relatively large) penalty in the kernel all the time.

For people whose process mix includes several processes doing such
access regularly, of course, the tradeoff goes the other way.

					der Mouse

			       mouse@rodents.montreal.qc.ca
		     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B