Capwap Mailing List: Re: Reserved bits and fields

[Bob O'Hara (boohara) (boohara]

Scott,

My concern is that requiring the sender to transmit zeros in the
reserved bits and fields leaves only one way to fix a protocol problem,
increment the version field.  Without the transmitter requirement, small
protocol problems found only after wide deployment might be able to be
fixed by using a previously reserved bit or field.

 -Bob
 
-----Original Message-----
From: Scott G Kelly [mailto:scott [at] hyperthought.com] 
Sent: Tuesday, August 15, 2006 7:03 AM
To: Bob O'Hara (boohara)
Cc: capwap [at] frascone.com
Subject: Re: [Capwap] Reserved bits and fields

Hi Bob,

Bob O'Hara (boohara) wrote:
> I believe that we are overspecifying the use of the reserved bits and
> fields in the draft.  For example, we place a requirement on the
> transmitter that these bits and fields MUST be zero ("All
> implementations complying with version zero of this protocol MUST set
> these bits to zero.").  However, we do not specify any behavior by the
> receiver.  I believe this is backwards.
> 
> Because the version field is available to all receivers of a CAPWAP
> packet and that version field will unambiguously identify all the
> reserved bits and fields, I believe we should change the requirement
to
> be that the receiver MUST ignore any value received in a reserved bit
or
> field.  This way, should a noncompliant transmitter send something in
> these fields, the receiver will not be confused by it.

I agree that we should say the receiver is to ignore non-zero reserved 
fields. This is consistent with the general IETF protocol philosophy of 
being strict in what you send, and liberal in what you accept. However, 
it is also common to require the sender to zeroize reserved fields 
(strict in what you send).

I'm not sure how/when this convention originated, but one 
security-related benefit is that it prevents the use of the reserved 
fields as a covert channel.

Do you feel strongly that the sender should be allowed to set reserved 
bits, or are you okay with zeroize at the sender, ignore at the
receiver?

Scott