[ih] XTP (was Re: Ken Olsen's impact on the Internet)

[Craig Partridge]

>     > As I understand it, the TCP effort had a close call with this same
>     > issue, when LANs started to be popular. I heard there was strong
>     > pressure to have a version of TCP tailored LANs but that Vint vetoed
>     > it.
> 
> I'm not sure quite which one you are referring to? Is this the 'trailer
> header' stuff from Berkeley? I'm not sure that Vint needed to (or, by
> then, had the capability to) stomp on any of this - it was pretty clear to
> most people that such things were a bad idea (and why).
> 
> There were also things like that attempt to design a 'hardware-friendly'
> transport protocol (i.e. one optimized for implementation in hardware)
> - was XCP the name? - and that also went nowhere, for similar reasons.

It was XTP -- Greg Chesson's believe that if the protocol was in hardware
it would be simpler and deliver data faster than the software implementations
that were always fighting with the network cards. (XTP = eXpress Transfer Protocol).

It has been a while, but my recollection is that there were some interesting
ideas in the resulting spec but it clearly was no simpler or faster.  The basic
state complexity was comparable to TCP, the impulse to optimize features
to the hardware/network added wrinkles that hurt, and there was no
innovation in congestion control that enabled XTP to send faster than TCP
over a link with unknown properties.

Thanks!

Craig