On Sun, Aug 03, 2003 at 03:50:16PM +0200, Martin Husemann wrote:
> On Sun, Aug 03, 2003 at 03:07:30PM +0200, Quentin Garnier wrote:
> > After all, the ISP only gets sessions from the L2TP
> > tunnel, with no distcintion if the client is using PPPoE, PPTP or PPPoA as
> > a transport to the BAS.
> 
> I don't buy the "with no distinction" part, and if it is true, this setup
> is braindead.

Actually, for the ISP it's always PPPoA, so the MTU is limited to the
ATM MTU - 8 from their POW (and they use 1500, as there's no point using
a larger one here). PPPoA becomes PPPoE inside the ethernet ADSL modem, which
isn't an IP equipement and thus can't fragment. The brockeness is at
this point (briging two physical networks with different MTU).
The ISP doesn't want to change the MTU on their LNS to 1492 because it
would impact all users not using an ethernet ADSL modem (USB or PCI modems,
ADSL routers). This ISP ships an ADSL router by default.

> 
> > So yes, it is a flaw in the design of
> > PPPoE in the sense that it cannot work peacefully with other PPP
> > transports.
> 
> The flaw in PPPoE is the small MTU (but it's hard to avoid in the general
> case, given what PPPoE tries to do).
> 
> I would very much prefer to have a PPPoA connection instead of PPPoE over 
> an ATM tunnel - and it's clearly a lack of clue at ISP/telco that forces
> me to use PPPoE at all - IMHO.

It's not a problem on the telco/ISP side, it's a problem with the customer
equipement. To avoid the problem, avoid ethernet ADSL modems.
BTW, does someone have an sbus copper ATM25 adapter that would work in a sparc
IPX for me ? :)

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 24 ans d'experience feront toujours la difference
--