RE: iSCSI: BASE64 and numerical values

[Paul Koning <ni1d@arrl.net>]

Excerpt of message (sent 21 May 2002) by Julian Satran:
> I am in the process of the reviewing the needs for large numerical values 
> (based on cryptography requirements for specified and other known schemes 
> that vendors may want to use). If The results will indicate that large 
> numerical values are of marginal use I will drop base64 for numbers. If 
> not I will leave it as is.

Diffie-Hellman and RSA are both based on arithmetic on large integers,
so clearly it is necessary to agree on which byte in the bytestring
encoding of those values is the most significant byte.

Then again, the same is true for non-numeric values: if you swap
around the bytes of an MD5 or SHA-1 cryptographic hash (as in CHAP)
things break just as badly.

> As for you "confusion and ambiguity in this realm of the spec (e.g.,
> decimal representation, leading zeroes, definitions of field size,
> byte swapping issues, etc.)" I am afraid that I don't see how base64
> different than hexadecimal and I assume a statement about ordering
> solves both.

True.

The remaining point is this: base64 is a standard encoding for octet
strings, and therefore for bignums.  It's not a commonly used encoding
for "normal size integers" i.e., 32 or 64 bit ints.

    paul