martha's comments

[[e-mail]]

Martha,
>I don't think that I ever said anything about BGP/IDRP and distance
>vector distribution graphs.  My reference to BGP/IDRP in the latest
>set of comments on the architecture document was an acknowledgement of
>your comment that loop suppression was taken care of by information
>contained in the distance messages themselves.

	In section 2.2.1 of the Architecture document you discuss Distance
	Vector approach with respect to distributing information. In this
	section you have rather confusing discussion on the subject
	of directed acyclic graph. If you think that this is a problem
	with BGP/IDRP then you need to explain it in details. Otherwise,
	what is the relevence of this discussion in the document.

>The "transient" loop problem revisited.  admittedly, i expect this
>event to be rare.  but unfortunately, jitter will not make the problem
>disappear.  X and Y need not be synchronized for this event to happen.
>all that must happen is that each receives and selects a distance
>vector (to Z) not containing the other before receiving the distance
>vector from the other.  if X and Y are closer to Z than they are to
>each other, then they are more likely to experience the problem.


	I would like you to explain to me how that can happen in a persistent
	fashion in presence of jitter.  I would suggest that we'll talk about
 it over the phone, since it would give us higher bandwidth.

Yakov