Jon Callas <jon(_at_)callas(_dot_)org> writes:
On 14 Mar, 2004, at 1:30 AM, Peter Gutmann wrote:
Explicitly allowing complex jumbles of packets seems to be just asking
for
trouble/interop problems, particularly when there hasn't been any
strong need
for them in the first 10 years or so of PGP's existence.
How about if I leave it the way that it is? The BNF does not permit anything
like a jumble of packets. Only literal packets can be in a stream.
That'd work, although there's still the current special-case of allowing a
string of literal packets. Is there any reason for this? My code doesn't
handle this, but that's because it never occurred to me that anyone would ever
construct a packet like that. It seems to be a clunky way to do
indeterminate-length packets (use a sequence of literal packets rather than a
single literal packet with indeterminate-length chunks). If there's some
reason for this, could the RFC include some text explaining it or discourage
people from using it?
(OK, I can dream up some imaginary scenarios for multiple literals where you
want to communicate, say two files in one encrypted packet, but standard
practice for that is to zip or tar/gzip them, not to use a multiple literal
packets).
Peter.