Wednesday, 19 May 2010

Green shoots?

Dead blogs of my friends litter the web. The roadkill of the information superhighway.
But green shoots can poke through at any time. Could this be one such place?

Need to find something to say.

Friday, 19 June 2009

I predict a battle for our (upload) bandwidth

Spotify is the latest internet phenomenon causing a Buzz. Hailed by some as the business model that could break itunes, it gives you the ability to think of song, and within seconds be listening to it for free - certainly intoxicating. And the collaborative playlists are a favorite of mine (great when four of you are sitting around with laptops in one room). Anyway, I digress: Spotify is on that cusp of just getting properly big, it makes it onto the news, everyone is seemingly getting an account - userbase is breaking out of the savvy minority (it only went open to all (UK) in February).

Yet Spotify is deliberately rolling out slow (at least relatively so) to allow smooth expansion and testing of the technology. Outside the UK, membership is still by invite only. But hidden away in terms and conditions is the p2p clause allowing Spotify to use your hardware to send data to other users, keeping their network overheads to a minimum. As yet, the user has very little control over the amount of upload bandwidth that Spotify uses (though there are plans to improve this). Even tech savvy users currently have a hard time controlling the bandwidth due to the viruslike way Spotify spawns port connections.

As long as Spotify remains restricted to a minority of users, everything will be fine. But I predict that, as with iplayer, if the userbase gets large, ISPs will start to feel the pressure. Spotify uses p2p which will quickly start to choke the limited upload bandwidth that the UK ADSL majority have (this is what the A=Asynchronous means). Unknowing Spotify users who don't even know what p2p is will start finding themselves subject to fair use policy restrictions. This was part of what eventually killed the BBC iplayer and 4oD download services in favour of (expensive) direct server streaming services.

At work, businesses will ban Spotify - it is probably already in violation of many company acceptable use policies but won't become an issue until more than a few are listening all day.

Spotify currently has a lot of goodwill, even amongst those that usually disdain proprietary software, simply because the product is so good. However bandwidth control and greater transparency to users who naively install the software is an issue that Spotify should address urgently if that goodwill is not to evaporate. And perhaps if p2p becomes the model for a greater number of mainstream applications, there will be an incentive for the industry to retune the A=Asynchronous a little, so that more of our new super broadband frequencies are dedicated to upstream .

Disclaimer: There is little original content in this post; it's just a collection of stuff I've been reading around. But I am off to start monitoring my own Spotify bandwidth usage. I live in a house with now 4 Spotify users, which might become a problem. At the very least I shall be setting the disk cache small.

Friday, 6 February 2009

John McCain

When remedies are past, the griefs are ended
By seeing the worst, which late on hopes depended.
To mourn a mischief that is past and gone
Is the next way to draw new mischief on.
What cannot be preserved when fortune takes
Patience her injury a mockery makes.
The robb'd that smiles steals something from the thief;
He robs himself that spends a bootless grief.

Thursday, 5 February 2009

A bit of snow

And the world grinds to a halt. More on this later.

The axiom of choice

The axiom of choice is incompatible with the axiom of determinacy.

The axiom of determinacy allows proof that ZF set theory is consistent.

The axiom of choice allows the creation of unmeasurable sets. There is no intuitive example of such a set, which is bad.

The axiom of choice allows the Banach–Tarski_paradox which is ridicululous until you remember that unmeasurable sets were ridiculous anyway.

The axiom of determinacy can be proved with infinite logic, which is cool.

So my vote would be against the axiom of choice. Sorry if that destroys your field but is there anything with a real application that requires the axiom of choice? Fourier analysis doesn't.

The joys of Wikipedia.

Friday, 30 January 2009

On H we shall stand or fall

Few professional groups in society have the privilege of self-assessment, but scientists in the UK have complained that the penta-yearly Research Assessment Exercise (RAE) is too onerous. So it looks as if in future we shall be judged mainly on our H-index.

Better get publishing then.

Thursday, 16 October 2008

The devil is in the detail

I've just taken a step back to write a brief summary of what I've done in the first year of this PhD thing, and what I hope to do before the end.

This exercise was all in the name of administrivia, so I decided to embrace it and had fun littering it with management buzzwords: "timewriting", "milestones", "progress review", "timescales", "project phases" and I even included a "gantt chart".

But actually it has left me reinvigorated (well perhaps that is a little strong) with what I'm trying to do. Amazingly I even felt inspired to start writing some thesis - it seems much more exciting than the devilish details which I have been fighting recently. These details have to be overcome, by nature of a PhD, but taking a step back helped me to remember that they are in the name of something bigger. I guess such details are what cause the second year blues.

I'm not quite sure why everyone in my department seems so vehemently disparaging of such management style stuff. In particular everyone hate progress reviews, transferable skills, professional development and free extra-disciplinary training courses. Whilst of course we all hate being forced to do extra things, some of them do serve a useful purpose, at least for me.