[rbridge] Consensus Check: Point to Point links

Thu Oct 4 21:28:36 PDT 2007

Eric,

Well, even if a consensus of the working group wanted to do maple walnut
waffle-cones with caramel, there is a problem that, as far as I can see,
that isn't in our Charter. But I think that this is related enough to
TRILL, as I will explain, that it would be OK to include in our output
if the working group wants to do so.

802.3-like interfaces for point-to-point use as indicated here can be
practical for Rbridge but not for 802.1 bridges because of the
characteristics of the TRILL protocol. In TRILL, the outer VLAN tag is
just an artifact for getting to the next Rbridge (or Rbridges) for TRILL
EtherType frames.  In a point-to-point link between Rbridges, that tag
and the outer MAC addresses are not logically necessary for TRILL
frames. Then tag can be omitted and the addresses can be a value ignored
on receipt.

However, whether you could implenent this feature is influenced by the
design of TRILL. For example, I believe the current protocol spec says
that local sources/sinks of frames on an Rbridge (for example SNMP) are
treated as if they were on a virtual port out of the Rbridge. That means
that, between two Rbridges, such frames area always encapsulated, even
if we has an Rbridge with a co-located management station talking to an
SNMP implementtion in a second Rbridge just one point-to-point link
away. If, in this case, such frames were sent in native form, it seems
to me it would at least make it a more difficult to use, for example, a
fixed MAC address interface or to correctly detect when the assumption
that the interface in point-to-point is incorrect. Thus, even if this
feature isn't mentioned in the TRILL effect, the difficulty and perhaps
even the possibility of implementing it could be affected by details of
the TRILL protocol.

There is also the question of whether the use of such an interface for a
link should be indicated in the Rbridge MIB, although if a fixed know
MAC address were used, I believe that could simply be checked for in the
MIB.

Donald

-----Original Message-----
From: Eric Gray [mailto:eric.gray at ericsson.com] 
Sent: Wednesday, October 03, 2007 3:10 PM
To: Eastlake III Donald-LDE008; Rbridge at postel.org
Subject: RE: [rbridge] Consensus Check: Point to Point links

Donald,

	Perhaps all of these things are true. 

	So what?

	The argument that a lot of people would like to do X could
be used to do a lot of things.  I'd like us to make maple walnut 
waffle-cones - possibly with caramel on top. 

	Ummm, does that sound good! 

	What does ANY of this have to do with TRILL?

--
Eric Gray
Principal Engineer
Ericsson  

> -----Original Message-----
> From: rbridge-bounces at postel.org 
> [mailto:rbridge-bounces at postel.org] On Behalf Of Eastlake III 
> Donald-LDE008
> Sent: Wednesday, October 03, 2007 2:25 PM
> To: Rbridge at postel.org
> Subject: Re: [rbridge] Consensus Check: Point to Point links
> 
> Hi Eric,
> 
> I believe the generally understood meaning of point-to-point is a link
> with exactly two devices on it, in this case RBridges.
> 
> It is true that one could always do this sort of thing outside the
> standards. But in Chicago the working group appeared to 
> believe this was
> sufficiently important to say in the standard and, while the consensus
> in the room was not that detailed, I got the general impression that
> people would be favorable to, in a later management document, having a
> MIB variable that would enable one to configure an interface as being
> known point-to-point, possibly point-to-point, or prohibited 
> from using
> this point-to-point outer address flexibility, or some such management
> configuration feature. There also seemed to have been at least some
> interest in an automated way of determining that an interface was
> point-to-point, perhaps based on the receive of 802.1AB frames all
> containing appropriate indications coupled with the receipt 
> of no native
> frames.
> 
> Point-to-point links are, in fact, very common in modern 
> 802.3 networks.
> 
> Donald
> 
> -----Original Message-----
> From: Eric Gray [mailto:eric.gray at ericsson.com] 
> Sent: Wednesday, October 03, 2007 11:11 AM
> To: Eastlake III Donald-LDE008; Rbridge at postel.org
> Subject: RE: [rbridge] Consensus Check: Point to Point links
> 
> Donald,
> 
> 	Given that it is not crystal clear what you mean by 
> "point to point link", I am not sure I agree with this, at
> least as you have worded it here.
> 
> 	If you mean that the link is a full-duplex Ethernet
> link and the two end-points have some definitive mechansim
> for determining that they are the only entities using the
> link between them, there may be issues with doing this.
> 
> 	For example, if the link is technically an Ethernet 
> link, then it is not unlikely that one or the other devices 
> may have multiple roles - i.e. - it may be both an RBridge 
> for some frames and either a regular bridge or end station
> (for example, a router) for others.  It's arguable that, in
> this case, the multi-role device is two (or more) separate
> entities - thus invalidating a "point to point" definition
> for this case (though only two distinct physical devices
> are connected via this link).
> 
> 	Without a clear agreement between involved entities,
> this sort of "short-cut" addressing is likely to result in
> higher-level (slow path) processing of many (if not all) 
> of the frames transiting the link for some implementations.
> Moreover, without an unambiguous determination of exactly
> when this would apply, it will not be unambiguously clear
> when a receiving implementation would have to switch to a
> "promiscuous listening" mode.
> 
> 	I believe omission of the outer VLAN tag suffers from
> the same ambiguity.  For instance, it is possible for two
> devices to have a "point to point" relationship within a 
> VLAN context that would not be the case without the VLAN
> context.
> 
> 	Hence it appears we would have to be explicit in what
> we mean by "point to point" link and how we expect that the
> entities (RBridges) involved would be able to disambiguate
> this p2p status for any given link.
> 
> 	If we are saying that two devices - using some means 
> out of scope for our specification - are somehow aware of an 
> unambiguous point-to-point relationship between them and can
> therefore use any MAC DA on transmission, and ignore it on
> receipt, we could make the same argument for virtually any 
> encapsulation choice we might prefer.  But, it would be as
> valid to observe that we don't need to specify what is - in
> essence - an out-of-context (mis)behavior between consenting 
> implementations...
> 
> --
> Eric Gray
> Principal Engineer
> Ericsson  
> 
> > -----Original Message-----
> > From: rbridge-bounces at postel.org 
> > [mailto:rbridge-bounces at postel.org] On Behalf Of Eastlake III 
> > Donald-LDE008
> > Sent: Tuesday, October 02, 2007 11:27 PM
> > To: Rbridge at postel.org
> > Subject: [rbridge] Consensus Check: Point to Point links
> > 
> > This is a check via the mailing list to confirm or refute 
> an apparent
> > consensus from the minutes of the Chicago meeting for a change from
> > protocol draft -05:
> > 
> >    If it is known that a link is a point to point link between two
> >    RBridges, then the outer header, if it is an Ethernet header, can
> >    have any source and/or destination addresses, those 
> addresses will
> >    be ignored on receipt, and the outer VLAN tag can be omitted.
> > 
> > If no particular controversy arises over this in the next two 
> > weeks, we
> > will declare it to be the working group consensus.
> > 
> > Thanks,
> > Donald & Erik
> > 
> > _______________________________________________
> > rbridge mailing list
> > rbridge at postel.org
> > http://mailman.postel.org/mailman/listinfo/rbridge
> > 
> 
> _______________________________________________
> rbridge mailing list
> rbridge at postel.org
> http://mailman.postel.org/mailman/listinfo/rbridge
> 