[Prev] Thread [Next]  |  [Prev] Date [Next]

Re: Sequence of attributes in a SCRAM message Nicolas Williams Tue Jun 16 11:00:26 2009

On Tue, Jun 16, 2009 at 02:50:24PM +0200, Arnt Gulbrandsen wrote:
> Nicolas Williams writes:
> >As for order within the SCRAM messages, I don't know the reason for 
> >nor history of that, nor do I much care whether there's order in the 
> >SCRAM messages. But I agree that having order _and_ attribute=value 
> >adds unnecessary redundance
> As far as I can tell, the order is necessary for the computer. The a=v 
> stuff makes scram messages easier to read for people, and it's only 
> about eight bytes.

It's possible to write a program that doesn't care about the order in
which the attrs appear.  That'd be programming 102 :)

> >-- one could have order and no attribute=, or attribute=value and no 
> >order, still have things work out.
> Yeah, but it would be more strain on the brain. 

We're having the age-old argument about redundancy.  How much redundancy
do we want in an on-the-wire encoding of a message?  ASN.1's BER's TLV
encodings have been problematic, in terms of bugs, meanwhile everyone is
always re-inventing PER and XML, which means that the debate never ends.

Let's not have this argument.  Instead let's agree that SCRAM is the way
it is and be done.  And keep in mind that several implementors have
prototypes already, or even code that awaits only consensus to get
deployed.  Changing SCRAM now for this little issue will cause those
implementors much hassle.