Re: [Idr] I-D Action: draft-ietf-idr-error-handling-03.txt

["Chris Hall" <[e-mail]>]

Brian Dickson wrote (on Sun 09-Dec-2012 at 20:20 +0000)
>
> What Chris is proposing is a negotiated option where the
> sender COMMITS to observing the specified order in which
> to send the attributes.
>
> Without that negotiated option, there is no method of
> influencing the sender's behavior. 

It would be straightforward for a BGP implementation to always use the
specified order, and always announce the "capability" -- this is
essentially unilateral.

The key thing is that receivers who understand the significance of the
capability can factor that into their error handling strategy.  (And
receivers who do not understand the capability can simply ignore it.)

> This is not to say that SOME senders won't use that order.

Indeed, in the absence of the capability, the receiver has to assume
that attributes may be sent in any order (as per current RFC)...

> However, any implementation that supports this new
> option negotiation, and either defaults to or exposes
> a "knob" for enabling it, will then need to use the
> SPECIFIC order to be compliant.

...or have some "knob" to force it to assume something else.

Chris