[e2e] on local ethernet throughput?

[RJ Atkinson]

At 09:55 29/10/01, Craig Partridge wrote:
>Modest defense of the kludgery.
>
>The carriers understand they have to deliver individual DSL lines
>back to any CLEC that resells the DSL service.  I.e. CLEC needs to be
>able to map something coming into their network as being from a particular
>customer line that they rent from the CLEC.  The carriers needed, therefore,
>some way to take a set of DSL physical lines and distribute them, flexibly,
>and individually to various CLECs.  They hit on ATM circuits as the way to
>do it -- not optimal, but not stupid either.

I'm told there is equipment (maybe from Copper Mountain) demonstrating
that a simpler, lower-cost approach would have been to use VLANs between
the ISP demarc and the customer.  Certainly some folks have deployed
roughly that sort of solution.  That approach avoids the PPPoE, 
reduces gorp, and avoids impacting the MTU.

Similarly, it is a pity that the DOCSIS standards don't allow for 
802.1Q VLAN tags and for VLAN-based downstream bandwidth allocations 
(the latter is just as important as the former) in the downstream
bandwidth allocation algorithm.  It didn't occur to anyone in the
DOCSIS process at the time.  It is hard to add that stuff now into 
silicon that's already deployed.

Ran