On 7/16/19 1:57 PM, Brian E Carpenter wrote:
The first one of all reads:
It could be that the
4 bits may be better devoted to making the flow id be a full
32 bits.
I think Brian's words of
SHOULD and MAY are important and a good idea too.
Frankly, this is positively insane. Either IPv6 is a variation
of IPv4, and shares the same ethernet type, with demux early
in the IP processing, or its new, and is demuxed at the ethernet
(and other) layers. Having it both ways is just idiocy.
So X is telling Y that Brian's idea is insane and idiotic, as far as
I can see. But IPv6 ended up with both its own Ethertype and an IP version
number field.
It sounds like the combative style of this 1996 poster was not effective
in persuading the working group. It may even have been less persuasive
than a more direct post explaining the disadvantages, for instance:
> Either IPv6 is a variation
> of IPv4, and shares the same ethernet type, with demux early
> in the IP processing, or its new, and is demuxed at the ethernet
> (and other) layers. If we try to have it both ways, we will
> waste bits on the wire and introduce processing complexity for
> implementators.
