Re: [manet] draft-ietf-manet-nhdp-10.txt - WGLC - Oct 5th 2009

["Justin Dean" <jdean at itd.nrl.navy.mil>]

While the idea of having optional "index" (they would have to be static and unchanging even if other interfaces went up and down unlike normal system index's) values and TLVs would help differentiate physically different interfaces to other routers it really isn't necessary.  The example I gave illustrates how NHDP would work with unnumbered interfaces without any changes.  The one caveat is that NHDP as is works with unnumbered interfaces as long as those interfaces do not "interfere" with each other, having two interfaces (w/ same IP) on the same channel wouldn't allow other routers to distinguish from which interface the hello message was generated.  As long as this isn't the case nothing needs to be added.  The only deployment situation that I can think of that this might be the case is with directional antennas, but even this would work if the physical interfaces of the same medium/address were arrogated together into a virtual interface.  (virtualizing multiple different interfaces, running on different media, into one WILL BREAK NHDP).  

An alternate proposal:
NHDP interfaces of a router which running on the same medium must be uniquely addressed.  
NHDP interfaces of a router running on differing mediums may be addressed the same (their "identifying," THIS_IF, address sets MUST match completely.)
NHDP may be run on a virtual interface comprised of identically addressed interfaces which are all running over the same medium.
NHDP cannot be run on a virtual interface comprised of differently addressed interfaces.
NHDP cannot be run on a virtual interface comprised of identically addressed interfaces which are running on differing mediums.

This allows for most deployments which I can think of without adding overhead or protocol complexity.  And it only adds complexity to those who would want to run a non "standard" network deployment and helps them do it in a way which would not break NHDP.  If there are deployments which others know about in which this would not allow NHDP to work correctly please speak up.

As I mentioned in my previous email the issue I saw was not with multiple interfaces with the same address but with differing routers sharing the same "identifying" address in their interface address list.  A simple fix is to disallow non-network-wide-unique addresses to be listed with the THIS_IF TLV value in a hello message.

Justin

> -----Original Message-----
> From: manet-bounces at ietf.org [mailto:manet-bounces at ietf.org] On Behalf
> Of Henning Rogge
> Sent: Wednesday, September 16, 2009 2:33 AM
> To: manet at ietf.org
> Subject: Re: [manet] draft-ietf-manet-nhdp-10.txt - WGLC - Oct 5th 2009
> 
> Am Mon September 7 2009 20:56:50 schrieb Henning Rogge:
> > Maybe I have an suggestion how we can add the "same IP multiple times
> > on different interfaces" without any overhead for normal operation in
> > the messages.
> >
> > new TLV (both message TLV and address block TLV):
> > IF_INDEX including a value of 1 bytes. The value must be 1 or higher.
> > (this supports up to 255 interfaces with the same IP, which should be
> > more than enough)
> >
> > ---------
> >
> > Every Node having the same IP multiple times add an IF_INDEX TLV as a
> > message TLV to HELLOS with with the duplicate IPs as interface IP.
> >
> > instead of working with "neighbor_interface_ip", NHDP will always use
> > a pair of "neighbor_interface_ip" and "neighbor_interface_index".
> This
> > will guarantee that the pair is unique. If no IF_INDEX is attached to
> > an interface IP, it's considered 0.
> >
> > All neighbor_interface IPs put into an HELLO will contain an IF_INDEX
> > TLV, unless the stored index is 0.
> >
> > ---------
> >
> > If we would run NHDP this way with no duplicate IPs, there will be no
> > change in the message structure at all. All interfaces will add "0"
> as
> > the default if_index to their identification pair, but this don't
> need
> > to be transmitted to neighbors.
> Just another thought about this.
> 
> If we add an interface index to the IPs in the hellos, NHDP and the MPR
> algorithm of OLSRv2 will be able to see what group of interface IPs
> belong to the same physical interface, which might help to do a better
> MPR calculation.
> 
> Henning Rogge
> 
> --
> Diplom-Informatiker Henning Rogge , Fraunhofer-Institut für
> Kommunikation, Informationsverarbeitung und Ergonomie FKIE
> Kommunikationssysteme (KOM) Neuenahrer Straße 20, 53343 Wachtberg,
> Germany
> Telefon +49 228 9435-263,   Fax +49 228 9435 685
> mailto:henning.rogge at fkie.fraunhofer.de http://www.fkie.fraunhofer.de
> GPG: E1C6 0914 490B 3909 D944 F80D 4487 C67C 55EC CFE0