[rbridge] Range of appllicability (was Re: TTL only - was RE: New fields in shim header?)

[Dinesh G Dutt]

Hi Eric,

Gray, Eric wrote:
> Silvano,
>
> 	The VLAN space you mention is larger in the RBridges case.
> It is possible to use 802.1Q encapsulation in the RBridge tunnel
> encapsulation - in a fashion similar to (but not necessarily the
> same as) the way in which it may be used in the native Ethernet.
>
>   
If I understand what you're saying, it is that we can add an additional 
.1Q tag between the outer L2 MAC address (carrying Rbridge next hop mac 
address) and the TRILL shim header and we can use the VLAN field in this 
tag to do what FTAG is being proposed for.

If this is right, let me list the reasons why I think FTAG is a better 
approach:
    - Using the VLAN field doesn't permit you to construct shared trees 
and use the FTAG as a tree identifier. Sure, you can redefine the VLAN 
field to be that, but then it's no longer VLAN as defined by .1Q.
    - I'm concerned about protocol overhead; this adds 32 bits to the 
protocol vs 10 bits by FTAG.
    - .1Q is a tag for Ethernet. If I want to use TRILL across 
non-Ethernet links, I need something that is independent of the L2 
transmission layer being used. While this case may not be the 90% 
scenario, it seems to intuitively violate layering in my mind.

Dinesh

-- 
We make our world significant by the courage of our questions and by 
the depth of our answers.                               - Carl Sagan