This thread is somewhat interesting but its quite apparent the majority
of this is based on opinion and/or thoughts that are less than accurate,
or possibly influenced by some insite gained in tests with a rather
limited scope. Try inserting a data scope at significant points of
interest between the end user's system and a few of their destination
addresses, capture the impact of changing MTU/MRU with various IP
applications and varying numbers of real users at the end user's
location, and we'll probably see a change of opinion on the subject.
There's no doubt about the MTU/MRU significance on selected vendor's
IP stacks (some of which have been pointed out), but most of the
comments relative to changing it have very little factual basis.