Tom Hume

Thank gawd for that... after nearly 2 weeks vanished off the internet, tomhume.org is back.

Hume's Conjecture - that all problems with any internet service can sooner or later be traced back to dodgy DNS - can once again be asserted.

Batty: I've done... questionable things.
Tyrell: Also extraodinary things, revel in your time.
Batty: Nothing the God of bio-mechanics wouldn't put you in heaven for.

One part of OverTheAir that I was particularly looking forward to was the hack-day element: getting a load of mobilists under one roof, working overnight within a tight timeframe to put together a novel demo. We'd had an idea beforehand that we'd like to do something which touched on the real world, and there's a fair amount of cynicism internally at FP about the value of some of the mash-ups you see out there today (which is not to say that there aren't some amazing ones too).

So Mr Hopper packed his soldering iron and Arduino kit, and we brought along a collection of devices: a GPS unit or two, a Sony Ericsson K850i and (of course) an iPhone. You can't hold your head up high these days if you don't do iPhone.

On actual ideas, I confess we were a bit fuzzier. I had some vague ideas of things I wanted to explore (e.g. bringing a bit of plausible deniability to digital communications, where it tends to get bumped out in favour of clear explicitness which is quite rare in meatspace), but had no idea of how to turn these into anything practical. So we went with doing something neat with robots.

The K850i is a strange device: a weird keypad and a mix of touch-screen and physical keys which a few handset vendors seem to be trialling at the moment. I find this hideously annoying, there's a really obvious jarring in my head when I press a button on these devices and have to work out whether it's actually been pressed by either looking for visual feedback or feeling through my finger - depending on which key I'm pressing.

But it has one lovely little feature, and that's an accelerometer accessible from J2ME via the mobile sensor APIs (JSR-256). Early this year one of our gold card days yielded an amusing little app using these APIs - a mobile spirit level - so we had a good idea of what we could do with them.

So robot arm, meet wavy mobile: we hooked up the accelerometer in the handset to the network, and (via a little bit of SSH tunnelling to work through the Imperial College firewalls at OverTheAir) connected it to a server running locally on a laptop, with said laptop driving the servos in the arm (connected using the Arduino kit). The result: with a 1-2s delay we could control the arm in at least two axes by moving the phone around.

This was quite fun, but I think we struggled to see it being hugely valuable to the human race. I had some vague ideas about getting the robot on stage, holding hands with it and swaying side to side singing "Give Peace A Chance", but they didn't generate the same excitement in the team that they held for me :)

So next step, add a camera: this was glued onto the end of the arm (making it quite heavy and giving us the sense of the machine straining slightly whenever it lifted up), and connected to an Ee-PC, which took photos from it at regular intervals and uploaded them to a publicly accessible web server. So now we had an arm which could be directed around, and which reported on what it saw: a little more useful, perhaps.

All this took us until about midday on the Saturday - an hour before the original competition deadline, with a few strained nerves as we ironed out last-minute problems and finally achieved consistent control of the beastie. It was here that things really took off as MarkNG and Bryan jumped in, taking the camera output and spreading it as far and wide as possible: in the run-up to 2pm (when the competition eventually drew to a close) photos started appearing on Flickr, in a Flash Lite client, on a Series 60 widget and, of course, on the obligatory iPhone. Our plans to show photos back down on the controlling MIDlet never quite emerged - some last-minute network problems courtesy of Orange (who suddenly started blocking our connections on port 8080 mid-way through the morning, necessitating a switch to Vodafone) had stumped a couple of us for an hour or so, and we never quite got round to completing this part.

The presentation was pretty ghastly, I'm afraid. Carrying a presentation laptop, controlling laptop, photo dumping PC, iphone, S60, iPhone, K850i, etc., all on-stage was a bit of a logistical problem (and we panicked a piece of kit would unplug/break/fall apart). Our preparation was minimal and I was far more nervous than in the previous days talk - when a few technical hitches stopped us showing the robot or phone on the large projection screens supplied, we called the demo short (after showing a film we'd made of it working). At least we managed to vividly demonstrate a theme Bryan and I had talked about the day before: that user interface concerns should be primary in any mobile project, and that development-led products are likely to suffer :)

I was genuinely surprised but obviously chuffed when we won the Best Overall Prototype; I (and a few other people I spoke to) had thought we were odds-on for "most over-engineered", but didn't expect anything else - and competition was strong. The prize, whilst generous, came second to hearing Matthew Postgate intone the words "Octo Bastard" in best BBC English - before the watershed! In particular I thought the LastMinute guys did a cracking job with their accelerometer-based sword-fighting game: I kick myself that we've never thought of this, it was so beautifully elegant and seemed to work really well. I hope that someone turns this into a real product, and I'm rather looking forward to an upcoming project where we will be collaborating with some of the folks from the team that built it :)

So, lessons from all this? Prepare more for presentations next time. Try and get some design involvement (both Hack Days I've done so far have led to the creation of bizarre hardware stuff, I'd like to go for elegance next time). And once you have your hack writing out to a web server, it's clear a whole load of other stuff opens up to you: widget platforms, mobile browsers, web services. As the mobile and fixed web converge (or the former subsumes the latter) it's the web-based services which are going to provide a basis for lots of interesting things, I think.

Also see Mark's writeup of the event, here.

So, I'm back from OverTheAir now, and just about caught up on sleep :)

I thought the event was absolutely excellent: a broad selection of topics for the talks (with 5 or so tracks running at any given time), an interesting selection of attendees (including plenty of folks from outside the usual crowd of mobile faces), and a brilliant atmosphere of collaboration. In particular the event seemed free from some of the operator-bashing which has a tendency to infect industry get-togethers; and I found the individual sessions to be particularly conversational (both those I attended and the talk that Bryan and I gave). It was really great to be able to discuss the realities of, say, mobile testing or approaches to mobile/web development with a crowd of smart people, all coming at the same problems from different perspectives.

In particular I enjoyed the Brian Fling talk, which managed to get even me worked up and enthusiastic about the way things are going (though I think it would be difficult for me, or indeed anyone, to equal the enthusiasm Brian shows for the iPhone).

The wi-fi was, I think, the best of any event I've ever attended: absolutely rock solid and snappy. The food was good, and the accommodation (which consisted of bean bags on a wooden floor) quite pleasant. Sleeping in my clothes for 3 hours before breakfast on Saturday morning left me a little... "fragrant", but what the hey :) And the few problems there were (the fire alarm leading to an evacuation on Friday) seemed to be passed over without trouble - a consequence of diligence by the organisers, I suspect.

I'll post separately about the two places where Future Platforms got most involved: the talk that Bryan and I gave about the PrimeSky astronomy project we've been doing for the Royal Observatory Greenwich, and our entry into the hack competition.

And I can't wait for the next event. Rumour has it there might be something hosted by the MoMo Amsterdam folks which could be interesting. Thanks to all the organisers - Dan Appelquist, Alex Craxton, Margaret Gold, Ian Forrester, Jo Rabin, and all the others I've left out.

...for the lack of service here. A confluence of bad luck: the DNS provider who hosted tomhume.org deciding to no longer host it, the company owning it not having any record of me, a trip in Berlin and to OverTheAir and my laptop being stolen.

Normal service restored soon. Probably by now, if you're reading this.

Andrei Popescu, Google Gears

Gears insulates applications from the network connection: they don't know whether they're offline or online.

Gears: OSS library with JS APIs. Local storage, expensive background computations (i.e. a basic threading model), allows access to local data using SQL. Google Docs uses Gears.

IE 6/7, Firefox 1/2 (3 soon). 1mb download on Firefox/MacOS.

Mobile is intermittently connected, slow CPU, little RAM, slow stable storage.

Gears web apps read and write to local store, changes queued for later synchronisation. Server communication is completely decoupled from the UI and happens periodically (perhaps a model for other mobile apps?). Interesting, this seems to partition all networked interactions into "storage" (async) and query-based (sync)... which I guess we have to do anyway.

Seems to have a read cache too ("resourcestore").

Then I got too hot and left :)

OverTheAir: Progressive Enhancements, Daniel Bodart

What is it? See wikipedia article. Not the same as graceful degradation: feature support not browser detection. Similar to TDD in terms of motivation and rewards: start with a dead basic service and add features which are made available where possible.

Text -> HTML -> XHTML -> XHTML Basic -> XHTML Mobile

The act of developing using this technique encourages better markup and accessibility.

Agile-ish in that you do the core stuff earlier, bells and whistles last. Early iterations deliver the most value. Sounds sensible.

Demo of E4 site, using the "behaviour" library. It's important to ensure that if your progressive enhancement fails, it fails completely without modifying the document messing it up.

Behaviour is to have the client make decisions, not the server. This helps deal with future devices which might come along without having to maintain a large database of pages. But it does mean that you're relying on lowest-common-denominator and the client being able to handle one basic version of your service.

My notes from todays talks - a bit briefer than normal, mainly because I've been paying attention to most of them...

Overall themes

A few years ago I remember mobility alone being justified as a reason to charge for mobile content. This notion seems to have been dropped, there's an acceptance that "internet thinking" is prevailing.

A few speakers have mentioned the idea that "all the enablers are in place": networks, handsets, etc.

Thomas Husson, Jupiter

iPhone is a positive trend but numerically weak: 200k sold in the UK, 90k by Orange in France, 80k by T-Mobile in Germany. iPhone users in France average 110mb pcm downloads. In the UK 60% of iPhone users download > 25mb pcm.

Web'n'Walk, the flat-rate T-Mobile tariff, has been around for 2 years and has 3m users: 5% of T-Mobile's overall customer base. Flat rate date might be clamoured for in our circles, but it's yet to reach the public at large. Low internet usage on mobile is driven by fear of cost and ignorance of benefits; 57% of consumers aren't interested in anything beyond voice and SMS (corollary: 43% want more). Mobile broadband won't hit >50% penetration until 2010 - in Japan is took 5 years to reach this rate too.

Users aren't prepared to pay for content, so it becomes ad-funded.

Triple digit growth doesn't mean much when you start from zero.

Click-through rates are high due to early-adopter audiences; they will come to track online rates. 47% of advertisers have trouble measuring their mobile ad spend. Now til 2010 will be a period of education; early growth in the mobile ad industry will occur after 2010. Ad spend isn't proportionate to attention given to a medium - TV gets lots of attention and lots of spend; print gets lots of spend and little attention; online gets lots of attention and little spend. In Japan ad revenue is roughly 2.5 euros per customer per year.

Antoine Couret, Bouygues Telecom

Bouyges have 2m I-mode users, 400 official sites, 50% of their I-mode users are active spending 3mb/customer pcm at a flat rate price of 9.9 euros. Previously PAYG customers used to download 1mb pcm.

Wow: only 3mb per month? That sounds v low and works out as 3 euros per megabyte - quite expensive for the user. Not that mb is a great measure of usage when a single video download can severely skew figures

Mainly email and "practical sites". They're now transferring elements of I-Mode to WAP handsets: flat fee, subscriptions for content, portal structure and free-to-receive email. PAYG customers have their bills capped. Content that's free online can't be premium on mobile.

Oliver Schmidt, O2 Germany

Not much that O2 has done with regards to data revenues has worked. This is an exception. They're the #4 operator in Germany with an unusually high 20% data ARPU.

If you price many separate services sensibly, customers still end up being very confused by a tariff incorporating many price plans. So O2 have 3 data plans: S, M, L.S is a PAYG package, time-based for occasional internet use. M is for 3G use on a handset. L is for laptop owners.

A new mobile portal has better UI, more prominent search, multi-column layout. Aims to generate more revenue beyond traffic and guide the customer to the wider internet.

Panel discussion: Gabor Nemeth (Magyar Telecom), Alexander Franz (3 Austria), Dan Rosen (AKQA)

Moderator: what's missing for a company like Coke to go for a mobile strategy?
DR: They're going for it already. Lots of others haven't, they're waiting to go in. There's a lack of good mobile marketing people. Ad inventory isn't as large as some big brands would like.
M: What are the other key enablers we need?
GN: Sub-laptops are important.
AF: As an operator, we're not just interested in selling megabytes.

M: What else would be needed from a brands perspective?
DR: Advertisers want common standards from carriers. We've been having this conversation for 6 or 7 years and it's holding things back. Flat-rate data obviously. iPhone has disproportionate impact - is it the device or the accompanying package of data? It's difficult to separate the two.

M: Why is something as simple as QR Code not simple to implement?
GN: For Hungary, standards aren't the problem. Hungarians don't read Yahoo or Google, they just use their own sites and services. These are our 4.4m customers.
M: But QR code doesn't mean "English speaking content". I hear contradictory things about capacity of the networks: 3 offer to air TV channels and do Skype over the network whilst others say TV and video could threaten networks. Do we have the capacity to grow the mobile Internet?
AF: Maybe for a GSM operator, capacity is an issue. But at 3 we have only a 3G network, so we'll use it as much as we can.

DR: Get things out there, test them, and then worry about monetising. It's possible to hypothesise too much. Top 5 mobile search terms in the UK are "drinking", "supermarkets", "mcdonalds", etc... completely different to the top search terms on the fixed web.

Tim from Youtube: how fragile are the networks when affected by large data traffic?
GN: When there are too many users in 1 cell.