Re: Clarifying the qualities we desire the DIGEST-MD5 replacement to have

[Simon Josefsson <simon@xxxxxxxxxxxxx>]

Hallvard B Furuseth <h.b.furuseth@xxxxxxxxxxx> writes:

> Chris Newman writes:
>> Nice-to-have features:
>> (...)
>> * Textual protocol like CRAM-MD5 over binary protocol that's harder
>>   to test/debug
>
> I don't get this.  SASL is not a protocol.  If a textual protocol _uses_
> SASL, it must turn SASL blobs into text - e.g. by base64-encoding them.
> If the SASL mechanism also base64-encodes, we get base64(base64(data)).

CRAM-MD5 uses hex encoding, so it will be base64(hex(data)).  I would
recommend against using base64 in the mechanism for this reason.  Base64
is considerably more error prone to implement than hex encoding.

This wastes a few bytes, yes, but I don't think the difference is
significant.

> If you are thinking of text vs ASN.1 I agree ASN.1 is harder to examine,
> but there are plenty of middle ways.  E.g. a fixed-field format with a
> field separator need not be textual.  And such binary fields could still
> start with 'token=' to identify them when you read the protocol.

Hex encoding the data as well doesn't cost much, and allows for simpler
string handling in some programming languages.

/Simon