[rbridge] Adopt draft-eastlake-trill-rbridge-fine-labeling-02.txt as WG document

[David Allan I]

HI Tissa:

Having two 12 bit tags IS a common scenario and defined in what was 802.1ad now rolled up into 802.1Q(2011). However they are not considered simultaneously, nor imposed simultaneously. My concern is that by requiring a double tag be inferred from a single tag and vice versa you are forcing a change to components in silicon that would be common for both Ethernet and TRILL. Same issue applies to applying filtering/pruning to the VLAN tree. 

My 2 cents
Dave

-----Original Message-----
From: Tissa Senevirathne (tsenevir) [mailto:tsenevir at cisco.com] 
Sent: Wednesday, November 16, 2011 2:07 PM
To: David Allan I; Erik Nordmark; rbridge at postel.org
Subject: RE: [rbridge] Adopt draft-eastlake-trill-rbridge-fine-labeling-02.txt as WG document

Hi David

TRILL allow VLAN based pruning. 6326bis draft already have added fine-grain labeling capability. This allow devices to prune based on fine label. Which in turn will limit the broadcast and ARP issue you brought up in point #3. 

On point #2. Having two 12 bit tags is a common scenario. Since we are limiting the interpretation of these two tags within the TRILL cloud and only within the cloud it is fine. Why ? Because TRILL link header is what 802.3 layer sees. Inner payload is for all practical purpose just like IP payload on a Ethernet encapped IP packet. Hence how the inner payload is treated should treated as a TRILL specific forwarding behavior not any shape violating 802.3.

Thanks
Tissa

-----Original Message-----
From: rbridge-bounces at postel.org [mailto:rbridge-bounces at postel.org] On Behalf Of David Allan I
Sent: Tuesday, November 15, 2011 6:49 PM
To: Erik Nordmark; rbridge at postel.org
Subject: Re: [rbridge] Adopt
draft-eastlake-trill-rbridge-fine-labeling-02.txt as WG document

I think I was the only person in the room to indicate disapproval for WG adoption....so I figure it is appropriate to offer my rationale...

It is a slightly expanded version of my comments minuted at IETF 81...

1) This is solving a TRILL design problem in the Ethernet layer.

2) It is non standard from an Ethernet point of view, and although implementations exist that may stack more than one tag, processing them as a concatenation of tags IMO is still likely new and inconsistent, so this is dependent on a proprietary and non-standard feature. Similarly inferring a tag stack from a single received tag is a new behavior as well as the re-mapping of the dual tag to single tag or untagged frame at the egress.

3) It actually has limited utility as it does not offer any improved broadcast containment beyond egress filtering as it is nested inside the rbridge header (assuming implementations can filter on a concatentated tag stack). So it adds no value in additional scoping for MDTs for flooding and ARP.

So I personally would suggest if you are going to require/mandate silicon modifications, they would be better off applied to the TRILL header itself vs. blurring the distinction between TRILL and 802.1 and introducing a discontinuity in standard compliant implementations. IEEE
802 has already provided a multi-sdo liaison a couple of years ago on how simultaneous multiple tag operations at a single hop is considered to be an architectural violation when proposals akin to this kind of abuse emerged during the GELS discussions.

So I have no issue with TRILL finding a way to support a larger VLAN space, I just do not agree with the proposed approach...it IS interfering with someone elses standard and how companies would implement that standard, and these days that is a really bad precedent.

I hope this helps
Dave

-----Original Message-----
From: rbridge-bounces at postel.org [mailto:rbridge-bounces at postel.org] On Behalf Of Erik Nordmark
Sent: Wednesday, November 16, 2011 9:14 AM
To: rbridge at postel.org
Subject: [rbridge] Adopt
draft-eastlake-trill-rbridge-fine-labeling-02.txt as WG document

At the TRILL WG meeting Tuesday there were several comments that the fine-grained label work is important and needs to be prioritized by the WG. There was also a rough consensus in the room in favor of adopting the document, as there was in Quebec City, but there were only about half as many people in the room that indicated they had read the document as at the last meeting. This is to confirm that it should be a WG document and ask people to read and comment.

   Erik and Donald
_______________________________________________
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