Jamie Zawinski wrote:
Why? This has been a constant thorn in my side. Does anyone actually
object to content-disposition, or is the problem just that nobody has
yet gotten off their butt and pushed it through?
(To be precise, content-disposition is currently an Experimental RFC)
Last I heard (about a year ago), the problem was internationalization.
Well, sort of. The specific issue that was raised was what to do with
the non-ASCII filenames some operating systems support. We found out
early on in the MIME development process that this is a can of worms.
Now, the size of this can of worms depends on how you measure can size ;-) If
you measure it in terms of the potential complexity of the entire problem
space, you'll find that it is pretty darn big. For example, filesystems
sometimes present different and inconsistent views of the namespace depending
on context. Trying to capture the precise semantics of this in order to
generate canonical filename parameter values s very, very messy.
But if you measure the size by how big a problem this is in practice, the can
actually turns out to be pretty small. For one thing, file name usefulness
correlates well with the degree of homogeneity of the environment. And people
adapt to this by making their name choices increasing generic the heterogeneous
the environment they're dealing with. So you found that while the problem
overall is intractable, solving enough of it to be very useful is well within
reach.
So what we ended up doing is essentially punting on the canonicalization
issue entirely. We now provide the tools for representing filenames that
contain 8bit characters, either with or without associated charset and
language indicators. If you want to do canonicalization, fine, and if not,
that's fine too.
Finally, given that this information is never anything more than advisory, I
don't forsee any substantive interoperability problems with this approach.
Ned