[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Endian orders

The most relevant and important point(s) made in this wonderful debate
(which ended up agreeing asserting everything has the semantics
of an ASN.1 OCTET STRING, and each data element in the recursive
list stream has a (BER-like private-class, multi-byte tag) label
to signal type information) are:

(a) there is no type information; its an octet-string, not an integer
(b) the octet string can be a programmatic representation of an integer
e.g. a tiny java applet so long as its tagged to signal that fact

It seems to me that there is a design consequence:

(c) the tag signals the context for interpretation of the octets, in
combination with the overall context established by the reduction process
for a given 5-tuple chain.

So a requirement is then, that:

It is clear then that the context signaling scheme
shall need to ensure tags are unambiguous for there to be a
deterministic result for determining type.

By analogy:

Just as, reduction algebra when performed over
an ordered n>2 length chain shall have "no state" (as discussed
in the IETF WG) so then simple tag interpretation shall need
to have similar properties such that typing semantics attached
to data by a validating-system shall be unambiguous.

At 05:22 PM 4/4/97 EST, Ron Rivest wrote:
>There seems to be enough of a consensus on network byte order, plus
>momentum for it in other crypto packages (including, as it turns out,
>the SDSI 1.0 implementation by Matt Fredette!), plus the suggestion of
>the chair for network byte order, that I think we should adopt network
>byte order and move on to more important things...
>Ron Rivest