[c-nsp] dot1Q trunk, not point-to-point

[Victor Sudakov]

Mike wrote:
> >
> > Is it required that a 802.1Q trunk is a point-to-point link between
> > exactly two switches? What if I have several switches with trunk ports
> > connected to a shared medium, should I expect problems?
> >
> > In my case, the shared medium would be a radio relay line acting as a
> > dumb switch which can however handle 1522 byte frames.
> >
> 
> Watch out - lots of wireless gear is outright treasonous and violates 
> blindly many aspects of 802.1d which come back to bite you in the butt. 

I don't think I will do STP on WAN links. Is it a good idea at all? All
redundancy will be on L3 using OSPF.

> This does not necessarily apply to 'real' equipment, such as licensed 
> band alcatel, dragonwave, and the like, usually it's the cheaper stuff 
> that's not on the market long that does it.

Hopefully this will be NEC's Pasolink. But if I'am unlucky and a
Micran http://www.micran.ru/english/ product is bought, all weird
things could happen. E.g. they declare 2 independent Ethernet user
ports, but they share a common MAC address table. You don't know about
it until you connect devices (or subinterfaces) with the same MAC
address to those two "independent" ports and find out that only 1 of
them is operational.

> 
> Some things I know about first hand include:

Thank you for the warnings, and thanks to all who replied, I'll look out.

[dd]
> 
> Using mac address translation to 'nat' mac addresses, and then using ip 
> inspection to simulate bridging... 

Why would anyone in their right mind want to do such a thing?

-- 
Victor Sudakov,  VAS4-RIPE, VAS47-RIPN
sip:sudakov at sibptus.tomsk.ru