ietf-822
[Top] [All Lists]

Re: Eudora VS MIME RFC

1995-02-10 23:30:46
"John W. Noerenberg" <jwn2(_at_)qualcomm(_dot_)com> writes:
Yes, I am frustrated that some implementors have against all
probability interpreted "unique" in such an interesting way.

This is not an issue of "unique"ness, this is an issue of what is
permitted in the body-part of a multipart.

The boundaries "-foo---" and "-foo-" are unique, but neither boundary
can be nested inside the other.  The result can't be reasonably
parsed by any grammar.

Brent Stilley <UCCXBRS(_at_)vm1(_dot_)ucc(_dot_)okstate(_dot_)edu> writes:
We just sprintf() the output of time and use that in the boundary. I guess
we could then reverse those bytes (with the obvious prevision for a tie-
breaker). But why should we need to?

You don't have to do anything nearly as difficult.  You could append
a '=' character, you could use a sprintf() format of "%08x", any
number of things.

You probably should use more entropy than just the system time, but
that's another issue.

This substring thing is absolutely counter intuitive! A reasonable person
would consider the whole boundary to be an atomic token. Comparison of two
tokens necessarily involves considering the length of each - not substrings!

tokens?  There is no tokenization step prior to applying the multipart
grammar.  The grammar applies to the raw octet stream.

Dave Crocker <dcrocker(_at_)mordor(_dot_)stanford(_dot_)edu> writes:
        For reference, I believe that everyone, including Steve D., is
wrong in their statements about 1521.  Really.

If you're including me in that statement, I'd appreciate your being
more specific.  Especially if you're taking issue with the examples in
my message with Message-ID: 
<cjCrPje00WBw0q450b(_at_)andrew(_dot_)cmu(_dot_)edu> on why
the 1521 multipart grammar "is just plain broken".

        People are mixing the terms boundary and delimiter as if they are
identical, yet the bnf uses them very, very differently.

Unfortunately, the nonterminal "delimiter" changes definitions between
1341 and 1521, in a way that is very significant to this discussion.
Alas, the text of 1521 sometimes mixes the two terms.

-- 
_.John G. Myers         Internet: jgm+(_at_)CMU(_dot_)EDU
                        LoseNet:  ...!seismo!ihnp4!wiscvm.wisc.edu!give!up

<Prev in Thread] Current Thread [Next in Thread>