[c-nsp] Problems sub-interface Catalyst 6500

[Brett Frankenberger]

On Sat, May 06, 2006 at 11:55:15PM +0200, sthaug at nethelp.no wrote:
> > > 	Asbjorn, as you appear to be walking reference for features, do you 
> > > 	perhaps know what does "phase 1" mean? Does it mean that we may even see 
> > > "real" subinterface support (in a sense what original poster in this thread 
> > > asked)?
> > 
> > 	My understanding is that this is on the roadmap for the future.
> 
> But the current hardware still only supports global VLAN space for the 
> normal "switch" type cards (excluding OSMs, SPAs etc).

Across the backplane, sure.  So each sub-interface will require an
"internal" VLAN allocated to it.  That doesn't necessarily mean that
you can't reuse 802.1q tags on interfaces, though.  Some line cards
already support VLAN mapping -- where the VLAN number on the interface
is different from the VLAN number in the switch (that is, you can
configure a trunk and specify that 802.1Q Tag 123 on that trunk maps to
VLAN 456 in the switch) ... with those line cards, today, you have have
two interfaces both using VLAN Tag, say, 50, but corresponding to different
VLANs in the switch.  The same hardware capability that supports that
could be used to support "real" subinterfaces (i.e. subinterfaces with
local VLAN numbering spaces, whith each such subinterface using an
internal VLAN allocation to carry packets internally).

The line cards that support VLAN mapping currently have limited mapping
capability -- some limited number of mapping entries, and groups of
ports for which the mapping has to be the same.  So they'll have to
develop better microcode, or the local VLAN space for subinterfaces
won't truely be "any VLAN ID on any port".  And it might not work on
any card.  (And, of course, I'm just speculating.  I have no
information that it's on the roadmap other than the quote above.)

     -- Brett