Re: I-D ACTION:draft-ietf-pppext-pppoe-mtu-1500-00.txt

John Fitzgibbon writes:
> 3) I understand the concern that allowing for a PPPoE header in the frame is 
> an IEEE issue, but, frankly, I don't want the hassle of taking this to IEEE.

That doesn't quite sound right to me.  I don't think the IETF should
be in the business of approving extensions to IEEE protocols any more
than the IEEE should extend things done in the IETF.

I don't think that ETOOHARD is a good answer here.

> A question: is this a suitable forum for allocating a new tag?

Since the original PPPoE documents were (I believe!) written in the
ADSL Forum and then submitted as individual submissions and published
as RFCs without (substantial) IETF working group review, I'd guess
that ADSL Forum (if it's still active) would probably be more
interested, but that publishing as an RFC wouldn't be entirely wrong.

(Ignoring for the moment my concerns about extending protocols that
are already in conflict with the IETF's stated architectural
direction, namely L2TP.)

> Some background on my rationale:
> 
> This is a real problem for me, and it's not performance related -- I have 
> residential DSL service and my provider, SBC, only offer PPPoE, (and they 
> seem *very* reluctant to change). Since I write network test software, the 
> lack of transparency resulting from an MTU of 1492 causes me some headaches. 

Right.  A lot of people are in that same boat.

Running plain old IP over Ethernet and using DHCP for both address
assignment and accounting, rather than resorting to tunneling IP over
PPP over PPPoE over Ethernet over ATM over DSL, might have been a
better idea.

> Given that I'm reasonably sure the hardware is capable of handling a 1500 
> byte MTU, the PPPoE limit seems overly restrictive. My thinking was that if I 
> couldn't change SBC, I'd try to change PPPoE.

Their equipment, as I understand it, terminates your DSL link and
sends the ATM directly to your ISP.  There it should hit an Ethernet-
over-ATM bridge and end up in some kind of PPPoE concentrator.

What's your plan for upgrading your ISP's equipment to handle the new
oversize frames?

> Upon investigation, (google), 
> it turned out that I was not alone -- many people have reported problems 
> arising out of the widespread use of PPPoE by broadband providers. 

Indeed.  It's a widespread problem with PPPoE.  There might be other
options.

> Presumably, there are many, many more who simply aren't aware that certain 
> relatively obscure connectivity issues may be related to the PPPoE link to 
> their service provider.

What about Vern Schryver's suggestion that you use RFC 1990 MP with a
single link instead?  Doing so would not involve inventing any new
protocols, and would involve changing fewer of the boxes in the
picture.  Just your system and the ISP's concentrator would need
standard Multilink PPP support.  It's possible that at least one of
these already supports it.

-- 
James Carlson, IP Systems Group?               <[email protected]>
Sun Microsystems / 1 Network Drive         71.234W   Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757   42.497N   Fax +1 781 442 1677

_______________________________________________
Pppext mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pppext