Re: [colorforth] DARPA takes aim at IT sacred cows

[kbk@xxxxxxxxx (Kurt B. Kaiser)]

Either this is garbled reportage, or the sources are uninformed.

John Drake <jmdrake_98@xxxxxxxxx> writes:

> Gibson cast some of the blame on the packet-based nature of Internet
> Protocol, which was not designed for foolproof delivery of
> messages. The protocol cannot guarantee delivery of e-mail, for
> instance.

IP is not designed to be reliable.  That's what TCP is for.  It will
keep trying until the message gets through.

> The packet network paradigm probably needs to change, Gibson
>said. Im not advocating throwing out the Internet Protocol
>completely, but we must absolutely have some mechanism for assigning
>network capabilities to different users 

In other words, users are prioritized and services allocated.  Fine
for the military, not so great for the ostensibly democratic internet.

>and that capability has to scale to large numbers of devices
>automatically. The commander wants to be able to send a message and
>have it delivered, completely, accurately and on time.

That's what IPv6 is for.  No need to re-invent, just adopt it.

> Another limitation with the IP approach is the inability to
>dynamically build networks. The military wants to quickly set up ad
>hoc networks.

This is not a packet issue, it's a routing, DNS, and link bandwith
issue.  You're either switched, packetized, or multiplexed, and I
doubt he's looking for a switched network or could multiplex an ad hoc
network without atomic clocks.  See the Roofnet technical papers.

Spread spectrum (802.11g) can work on wireless networks.  It's still
packetized.

> Static networks are no good for tomorrows battlefield,
>because everything will move around all the time, Gibson
>said. What we need is dynamic scalability. Todays networks
>are stationary and have a static infrastructure that provides service
>to static end-nodes. Moving the node outside its standard service
>area requires reconfiguring something. Moving infrastructure always
>means reconfiguring something.
>
> As a result, DARPA wants to fund development of new protocols or
> enhancements to the existing IP that will allow nodes, such as
> computers, to automatically sign on to networks in their vicinity.

Except for the military, this is a commercial/political issue.  The
network must let you in and somehow payment has to be handled.  Since
NAT isn't going to work, IPv6 is essential.

> Another aspects of the networking that DARPA wants to revise is the
>seven-layer OSI stack, long held as the basic foundation for building
>network protocols.
>
> The OSI model was not designed for wireless communications devices,
> said Reggie Brothers, a DARPA program manager.
>
> The OSI model served us pretty well for the stable, predictable
> world of wireline communications, Brothers said. Mobile
> networks are nothing like that. They are unpredictable and highly
> variable. We need to think of different layers of the stack to relate
> to one another directly, like a mesh, instead of one level up to the
> next.
>
> The increased complexity of the network stack would let nodes enter a
> network quickly and without human intervention, Brothers said.

*More* layers is a solution??  Or does he mean turn it into a 
big, unmaintainable hairball?

> The von Neumann architecture will also come under scrutiny from
>DARPA.
>
> It is time to ask the harder questions about the ways of computer
> architecture weve been using for the past 30 years. Is it time to
> scrap the von Neumann architecture? asked Anup Gosh, program
> officer for the Advanced Technology Office.
>
> This architecture, which defines the basic essential parts of a
> computer as the processor, control unit, memory and input-output
> devices, has been used as the basis for design for nearly all
> computers built since the 1940s.

That's because it works.  We tried parallel computers a decade ago and
nobody could figure out how to program them for general purpose
computing.  e.g. Thinking Machines, out of business.  The day will
come: today's computers are already not strictly von Neumann because
there are multiple execution units "behind the curtain" of a Pentium,
and SMP and distributed computing is common.

This is of course an issue for Chuck's multiprocessors.

> One of the limitations inherent in this approach is that when an
>application malfunctions, it can affect other programs, Gosh
>said. Program bugs also are vulnerabilities that can be used by
>adversaries to attack the entire system. What military networks need,
>Gosh said, is a way to isolate software programs at the hardware
>level.

Heaven's sake, that's what an operating system kernel is for.  Too bad
he's running Windows, like the rest of the Armed Forces.  It gives a
bad impression of how secure a properly audited kernel can be.

Either he means have several independent computers at the "hardware
level" which communicate with each other locally over authenticated
links, or he wants to move some more functionality from software to
hardware so that it is tamperproof, or both.

The opinions which are reported are more applicable to a dedicated
(wireless) military network.  They don't have much bearing on the
internet (hopefully).

I expect the military to require large numbers of small, autonomous
entities communicating at fairly high BW.  Sounds like a job for
dedicated hardware, i.e. Forth chips.

-- 
KBK

---------------------------------------------------------------------
To unsubscribe, e-mail: colorforth-unsubscribe@xxxxxxxxxxxxxxxxxx
For additional commands, e-mail: colorforth-help@xxxxxxxxxxxxxxxxxx
Main web page - http://www.colorforth.com