On Tue, 2008-04-08 at 11:17 -0400, Jeffrey Hutzelman wrote:
Alexey Melnikov <alexey(_dot_)melnikov(_at_)isode(_dot_)com> wrote:
Is this reasonable? I think we should at least prohibit other ASCII
control characters.
I think it is reasonable to require the character to be printable.
e.g., we don't want Unicode BOM in there, either.
Isode's implementation is even more restrictive, as it disallows ¬, :,
/, etc. Should they be allowed?
There are certainly some classes of characters, such as letters and digits,
that all implementations should accept. But there also should be some
flexibility; for example, some implementations may find it useful to
disallow some subset of [][/\\:;,] or other characters, depending on the
underlying OS.
please let us not go down that route, the list of potentially
problematic characters is endless. an implementation can easily use,
say, a URL-style encoding if it uses a filesystem as its store.
I'm in favour of setting aside one character for hierarchy purposes
though, e.g. "/", even if it has no meaning today.
--
Kjetil T.