Re: NHRP specification comments

[David Horton <[e-mail]>]

> |Jim,
> |
> |	Some more comments (specifically relating to operations):
> |
> |All references are to draft 8 (beta).
> |
> |Section 6.2.1 ("Caching Requirements") - in the paragraph relating to
> |caching at "Transit NHSs" - currently this says that cache information
> |MUST be discarded when it has exceeded its holding time.  It also says 
> |that this information may be used to form a reply to a non-authoritative 
> |request. 
> |
> |I believe that the negative effects of caching at intermediate NHSs can 
> |be substantially reduced if hold time returned in a non-authoritative
> |response MUST be the result of decreasing the original hold time by the
> |elapsed time since initially caching the information.  This information
> |is either immediately available or easily derivable (depending on your
> |implementation) and limits the length of time that stale information can
> |persist in the system to approximately that intended by the original
> |responder.
> |
> |I would be interested in whether or not this is a commonly-held opinion.

That is what our implementation does.


> |
> |Section 6.2.2 ("Dynamics of Cached Information") contains a note (at the
> |end of the subsection entitled "NBMA-Connected Destinations" that might
> |be interpreted as meaning that it is okay to cache information contained
> |in a Next Hop Resolution Request (as opposed to Reply) because it is 
> |considered stable.  Is this what you mean to imply here?

The source address of the station making the request is presumably
reliable, so in the absence of overriding static or formal register
messages from the station to the NHS, the data can be cached.
It can't be considered authoritative because the client won't
be sending purge messages to any other addresses than its home NHS
for example at shutdown.

We must also allow the formal datato override this informal data
as the device may be sending on multiple ATM addresses at the one
IP address, and hence have a preference for one address for some reason.
(Perhaps some SELs are used for different QoS traffic types).


> |
> |--
> |Eric Gray
> |


-------------------------------------------------------------------
David Horton
Centre for Information Technology Research
University of Queensland, 4072
Australia

Email: [e-mail]           Phone [phone]     Fax [phone]