Wednesday, May 28, 2008

making a multiple hash of it

so just re-reading this really nice classic paper
Probabilistic Counting Algorithms for Data Base Applications (1985)
Philippe Flajolet, G. N. Martin, and realized that one of the tricks they use
is to do multiple hashes of the key (they then count the occurrences of bits
in a neat way to get the cardinality - very cool - so how does this relate to
Bloom Filters which count the approximate set membership, and to CAN's, which use multiple hashes to distribute keys in a search space as uniformly as possible, or even to coordinate estimation systems? Seems like there's a small monograph in this...

Monday, May 26, 2008

internet research for kids

I've been on sabbatical in the last year visiting lots of labs and
talking about research directions and one idea that came up is a
reaction to something I think and I believe you and dave clark have
also said in the last couple of years about academic CS research
(at least Internet related CS research) having gone awry.

The pressure to look at short term fast return has pushed many
academics into effectively competing with industry R&D - this means
that many papers are increasingly being produced that look like
startup white papers (often are) and do not have the depth of work
behind them that one might expect from a core CS conference (let
aloine journal) 15 or 20 years ago. This is not universally
true (self promotion - I believe we published first papers
on work on Xen after about 9PhDs and 3 years of faculty work,
and on Xorp and Metarouting work were both in similar ratios and have
strong underpinning ideas).

What my colleague Tim Griffin refers to as the Hotnetization of
communications research has happened.

Part of the problem is that really smart people get a big buzz out of
being creative, and only some buzz out of the sweat to really nail the
details. But the details matter. One solution to this would be a big
cultural shift back to long term funding and rewards to academics for
long term (not for short term) work - that isn't going to happen in a

Another solution is to provide a competitor for short term idea
generation and demonstration to force academics out of the short term

SO here's one idea:

build a system for kids to write new internet scale applications.
give them a sandbox (planetlab junior) to deploy these.
Think lego-mindstorms for the internet.

A whole bunch of things like facebook, myspace, skype, bebo, flicr,
IM,, are totally obvious to kids - 10 year olds would have built us
prototypes if they were given the simple tools to create
distributed applications with simple GUIs I think they would
create 1000s of them - it would take an organisatio nlike
Cisco (and/or microsoft) with a scools outreach programme,
and some resources (not a lot) - it would be nice to
incorporate cell phone handsets in some simple way.

There are some nice example languages (as above, lego
mindstorms, but also Alice (from CMU) which could be extended
easily to do this....

just a thought!

Thursday, May 22, 2008

biometric - dont let the tech speak bamboozle you

governments (well the british government) likes to pretend it is up-to-date and techie, when in fact we all know that thatcher, major, blair and brown are all pretty technophobe and so there's no real comprehension at the top of the food chain, and since most civil servants train at oxford doing PPE (politics philosophy and economics) rather than at Cambridge doing NatSci (natural sciences - or medicine - where people get jobs in the Real World), the mandarins of White Hall are pretty useless - this is why they keep proposing idiotic things like ID cards, the NHS IT programme, and biometric passports (and the reason the write such bad RFPs and get no-one bidding is that anyone with an ounce of real business skill can stand looking down the river east of the houses of parliament and see a 100 places where they can get a job with 10 times the power and 25 times the income.

so. biometrics. what's that about then?

well, a photo is a biometric. bio - life. metric - measure. so its like a piece of information gathered about you hopefully something a third party (not you or the person who made the passport) can study and compare with the True Life version they can get from you - this then means that they trust who you are because they trust the person issuing the passport checked who you are and only gave you the passport since the other info (birthday, place, citizenship etc) were entitled to, and that the passport cannot be fiddled with (affordably) and that you can't be fiddled with (affordably).

so. basically, voodoo then. I mean who really understands DNA or Iris scans or finger prints or retina scans ? who understands encryption? who understands challenge response protocols or redundent coding of information?

well, I suppose we do. But we don't work for the government, do we? so why do you believe they'd get it right?

but it is all voodoo, because basically the government likes things that sounds technical (remember ghost busters?: "back off: I'm a scientist", and "it's technical".

No, next time someone says we must have biometric NHS entitlement ID cards, point out
1/ bio-washing powerder has been shown to cause rashes in some people, so how do we know bio-metric cards wont casuse rashes? have they been tested? have the RFID readers been checked that they don't cause cancer?
2/ we havn't gone metric in the UK yet in all things (e.g. road distances, measures of beer) so why should we go metric in identity? We still have feet - we do not walk about on iambic pentameters. We have hands and measure horses with them - indeed, we have cars that are powered by horses too, even if the force is also measured in Newtons.

That will shut them up for a while.

Friday, May 16, 2008

laptop v. smartphone

so i've been traveling a lot recently and carry a v old (crusoe) based vaio small lapto p- it is about 7-8 years old and runs win ME or Linux 2.4.17:)

I also carry an HTC touch

the security folks in airports always ask me to take out the laptop - occasionalyl they dont notice it, but they never ask about the phone - the phone has a faster processor, 4 times as much memory and a faster wireless interface....and about 6 times the battery life..on the other hand it is 7 years newer (and 4 times cheaper)....

luckily, i only have to carry one charger for both:)

once again, the staff in the aircraft and airports annoyed the *** out of me by using phrases like "once again" when they havnt said anything before, and also "last remaining" when they mean "remaining" or "last"...

it took me 9 hours to get from stockholm to london last night due to the airtraffic control being down in amsterdam - distributed systems (what is the definition of a ...) hah!!!

Monday, May 12, 2008


Not many people know this, but if you run one of those visualisation tools
on the xen source tree, and on the AS topology of the internet,
at any point in time over the last 5 years,
the result is identical.

this forms one plank in my conclusive proof
of the non existence of Richard Dawkings

Saturday, May 10, 2008

energy saving gadgets

so i have a lot of legacy technoogy in the house which gets left on by certain small people - tvs, games consoles, hifi etc etc

so of course it'd be nice to have it all gracefully shutdown when people are not in the room - how cheaply could one combine motion sensors (say) with timers and mains adaptors that just did this?

physcally hierarchical/nested flash memory - matryoshka usb sticks?

so say i have a 5 year old 128M usb memory stick , and i have a 2 year old 1G and a 1 year old 8G - why can't i stack them? couldn't i "write thru" - so I plug the newsest biggest one into the USB slot and it has a microusb conenctor on the bac and i plug the middle one in there, and that has a nano-usb connector, and I plug the oldest smallest one in there:

laptop <=usb 2===>8G stick<===microusb===> 1G stick <=== nanousb===> 128M

why? trade off in write time v. capacity?
redundency and so on...

marketting - easy- just paint them with freiendly russian traditional scenes like the dolls

Friday, May 09, 2008

shall i compare thee to a lewd cuckoo?

there i was cycling along the backs
composing a three part harmony round
with a libretto lavishly drawn from
shakespeare, anglo saxon folksong
and jazz (like to the lark,
at break of day arising
every day we sing cuckoo,
but how strange the change
from ursa major to minah-
bird brained we sing...)
and, it being a beatiful sunny may day
in fenland
I get a mouth full of greenfly... i cannot remember
the tune

Thursday, May 08, 2008

naming conventions (for research projects)

so i just posted an answer on the blog to where the name (abbreviation) xen came from - while we are there, metarouting (tim griffin's fine project to replace hand crufted routing protocol code was a local idea too...this project is not quite at the take-over-the-world stage, but it is getting there with recent heroic implementation work. A related work is mark handley (et al)'s XORP open source router project. Although the metarouting gang switched to using Quagga for some pragmatic reasons, other projects (e.g. with Lancaster) have combined xen and xorp for virtualisation of routers - indeed, many US people are trailing this work (for a change:)

also fun is david greaves' work on systems that are corect by design - see

his blog on current work on tools and methods