Anyone tried 1500 MTU on Infinity?

[watercooled]

Just wondering, has anyone had any success using RFC4638 'mini jumbo frames' to preserve 1500 MTU of Ethernet frames with BT Infinity? According to some places, it's fully supported by the network but strangely doesn't seem to be used by the HH3.

I'm currently on VM so can't experiment myself, and the net is full of contradictory arguments often polluted with confused posts.

Reason for asking, path MTU discovery isn't enforced with IPv4 so some places probably don't correctly implement it, leading to fragmented packets and unnecessary overhead. That, and I'm just curious.

A report from the MTU section of the following test would be interesting too.
http://netalyzr.icsi.berkeley.edu
For reference, VM shows 1500 both ways, no fragments.

[Dan-M420]

I think it would be best to keep it as it is, BT infinity is still DSL to a point, (Unless its the fibre to the house thing)

[watercooled]

That's irrelevant, ADSL generally uses ATM on the wire, VDSL (Infinity) uses PTM*. Besides, MTU is dependant on the the PPP endpoint supporting RFC4638, which according to a few places Infinity *does* support. For example, check under 'technical' on this page: http://www.aa.net.uk/broadband-FC.html

*In theory, ATM, or more specifically AAL5, can carry 1508 octet or larger packets. Any upper packet limit will be because of ISP RAS settings, and unless I'm mistaken is to avoid having to use and pad out an extra ATM cell for every packet, increasing overhead further on the DSL connection. AFAIK, there are no traces of ATM left with Infinity.