Posts by Neil Davies 1

How fast is a piece of string? Boffin shoots ADSL signal down twine

Not necessarily - they haven’t yet claimed they can get a working PPP session over string-net yet

Report: Underwater net cables are prime targets for terrorists and Russia

Forget people - worry about Mother Nature

When i get involved in planning this sort of thing, it is the mud slide due to something like a mid Atlantic earthquake that is the greatest concern.

Correlated failures which generate excessive offered load on the rest of the infrastructure - a lot more disruptive, but not terror?

Multics resurrected: Proto-Unix now runs on Raspberry Pi or x86

Re: There were five in the UK..

One of its main functions was to front end the cray - I installed the coloured book network protocols on that one. AUCC wrote the protocols

Manuals

I've still got a bunch of them - Multics Programmers Manuals - got to dig them out

Source code

Yes - we did microfiche and tape, we could mount the source on a separate logical volume

Net neutrality: How to spot an arts graduate in a tech debate

Re: Proof

Take a look at the report: Glasnost approach would detect the presence of such differential treatment, but where on the supply chain is the issue?

There never (rarely) is a single "domain of control" that is responsible for the delivery to you - and that's Ofcom's issue, if you can't isolate you can't enforce.

In the US it is slightly clearer (and remember this is where the stuff started) but even there there are other issues of responsibility, in the UK our digital supply chain is much more involved.

Disclosure: I was one of the report's authors.

Re: Stochastiv vs. deterministic

Agreed - that was a "typo" due to journalistic deadlines - has been updated.

Technically stochastic is not really non-deterministic (that means you know nothing!) - it expresses levels of uncertainty in the evolution of the system. Interestingly it does have analogies with chaos, but network protocols can take a stochastic system and make the outcomes become more determined (i.e. keep the system within its predictable region of operation and, for example, make the time to complete of a large file transfer more predictable) or, it the presence of nastier forms of non-determinism (which is actually what happens in today's networks) the emergent properties can vary widely - just as you experience.

Disclosure: I was one of the report's authors.

That's not precisely what the report says.

It says that a) if you suspect differential treatment you can't localise it to where in the delivery chain it occurred with the current tools [hence not "fit-for-ofcom's needs" - who do you chase after and how?]; b) there are lots of differential treatments that can't be detected with the current tools; c) The differential treatment is the wrong question - you can test for a quality floor much more easily.

All those scientists who look for "signals" in randomness have models for the signal they are looking for, there isn't a model to test against at present (and "net neutrality" is an aspiration not a model)

disclosure: I was one of the reports authors.

Web daddy Tim Berners-Lee calls for net neutrality in Europe

Sorry Sir Tim. You need to get your science right, this sort of packet treatment leads to uncontrolled performance collapse for everyone. The science says it, so does the practice.

Network Neutrality is far too weak a goal, it doesn't meet the end user requirements (which is for applications to work) nor the network operators goals (to create value chains from which they can extract a reasonable reward for their capital and effort).

EU delays vote on roaming, open internet after TRANSLATION fail

Next π has the value 3

They don't get statistical multiplexing do they? Even a FCFS tail drop queue is traffic management!

They *really* don't want the world in which it is a complete free for all - the resulting network dynamics (and costs) are not the ones they thought that they would get...