Re: Comments on the L2 VPN framework and solutions documents

Cheng-Yin Lee wrote:
> Mick, Matt,
> 
> Mick Seaman wrote:
> 
>>From the point of
>>view of control connectivity the pseudo-wires have to establish a full mesh
>>that is never perceived by layer 2 protocols as being less than complete.
> 
> How can this "lower layer emulated LAN" (using full mesh split horizon)
> be perceived by IEEE L2 as being _never_ less than complete? 
> 

One has to take precautions/measure to ensure that the virtual port to 
the emulated LAN is operationally down if there is an incomplete mesh 
for that emulated LAN.  There are various ways this could be accomplished.


> 
>>But ensuring the
>>emulated and real technologies are transparent to the higher layer, one
>>can build arbitrary networks based on traditional bridging models.  I
>>hope thats a good thing.
> 
>>From the architecture point of view, it looks clean to make the lower
> layer transparent and to mandate the lower layer should do its job
> properly (in this case IEEE 802.1ad assumes PPVPN VPLS always provide a
> properly emulated LAN, no configuration loops, connectivity issues,
> etc). However, I do not know how loops in the forwarding paths can be
> prevented, if a tree is concatenated/spliced with the lower layer
> forwarding path (or another tree), unless all common nodes in the
> forwarding path (both higher and lower layer) participate in one common
> spanning tree or a similar algorithm to prevent forwarding loops.
> 

I'm having a hard time parsing that last sentence.  Loops in a VPLS are 
prevented with split horizon, and I'm not sure what tree is being 
spliced/concatenated with what.  But emulated LAN segments would be 
bridged to other emulated or real LAN segments, and if the bridged 
topology is not a tree, then STP has to be used to reduce it to a tree.

- Matt