Jonathan Stone said:

>Although it's unicast, the closest analogy I know of is NETBLT.
>AFAIK, the authors of NETBLT have never released the implementation.
>A long-term colleague of those authors once told me that nobody with
>access to source ever has or ever will allowed it to be deployed,
>because nobody wanted to be known as the person who _completely_
>melted down the Internet.

Long ago and far away I used to work for a company called Charles River
Data Systems. During 1987-19889 Charles River UNOS realtime UNIX was ported
to the then new Intel 80386. One offshoot was a project to build backend
databases accelerators for Vax/VMS systems running Oracle. The "accelerator"
backed was connected to the Vax via a private 2 node ethernet network. 

The two main developers on the project came up with a non-TCP networking
protocol that looked a lot like RFC 998. It gave them a 30% or so increase
in throughput over TCP/IP and CRDS UniverseNet. In the end they went with
TCP/IP, because TCP/IP was an important checkoff box on government contracts.
(As opposed to the evil DECNET) :-)

Mind you, that was a long time ago, and it is doubtful that any of that code
lives. When they wrote that code there were about 500 systems hooked to the
internet, with a lot of other systems using UUCP. The code never got used on
a shared network.

Meanwhile, today the internet is like trying to suck a small, thin stream of
pure water through a large pipe filled with sewage.

-- 
Rick Kelly	rmk@rmkhome.com
		<http://www.rmkhome.com/>
		<http://rkba.rmkhome.com/> - the right to keep and bear arms
		<http://wolf.rmkhome.com/> - firearm forums