No Content, No Fuss: Category research

Melange hits the euro-spotlight

anil@recoil.org (Anil Madhavapeddy) — Sun, 01 Apr 2007 23:12:00 +0100

It's been a busy old March (release management of the new XenEnterprise sucked up most of it). I did take a break and go over to EuroSys 2007 in Portugal to present the language and compiler I implemented as part of my PhD work (read it here).

Although the talk I gave was was a bit underwhelming (more preparation and time-practise next time!), I met a whole bunch of really interesting people. My argument about rewriting whole applications also didn't get laughed out the room as I thought it might, as people recognise that retro-fitting safety enhancements on existing languages is a bit of a dead-end road to go down. It has definitely inspired me to make more time to spend on polishing up the Melange applications for a proper release in 2007.

In a pleasant surprise, it also won the Best Student Paper award of the conference as well!

Deens, welcome to the Internet!

anil@recoil.org (Anil Madhavapeddy) — Sat, 30 Dec 2006 01:11:00 +0000

Inspired by finishing my PhD corrections (!) today, I decided to hook up the DNS server from our Melange project up to the Internet. The authoritative server is called deens (since the co-author is one Tim Deegan, geddit?), and is written in pure OCaml.

This is all rather experimental, to put it mildly, but I stuck in the zone file below, hooked it up as a delegate to our main name-servers, checked it against the DNS Report, and it all seems to be working!

$ORIGIN deens.recoil.org. ;
$TTL    240
deens.recoil.org. 604800 IN SOA  (
    deens.recoil.org. anil.recoil.org.
    2006122401 3600 1800 3024000 1800
)
        IN  NS     ns1.deens.recoil.org.
        IN  NS     deensns.recoil.org.
ns1     IN  A      194.70.3.132
dynamic IN  CNAME  dynamic.recoil.org.
static  IN  CNAME  static.recoil.org.
anil    IN  CNAME  dynamic
stats   IN  CNAME  dynamic

I also modified stats.recoil.org to be an alias to stats.deens.recoil.org, so all the requests for that domain will go via the deens setup. You actually need a user/pass to access the site, but that doesn't matter; if it gets that far, the DNS bit has worked.

There's still an awful lot of tedious work to get the server into a production-ready state, such as proper logging, more error handling and recovery, etc., but I really hope to find the time in 2007 to polish this up somewhat. Performance is excellent already; faster than BIND by quite a lot, and it can optionally use more memory to cache responses to shoot up to crazy levels.

Incidentally, the dig replacement utility also seems to be working fairly well, and David Scott has been messing around with a Bonjour implementation that will get finished sometime in 2007 as well (honest!).

Virtual virtual virtual fragging?

anil@recoil.org (Anil Madhavapeddy) — Wed, 30 Aug 2006 22:35:00 +0100

An oft-cited criticism of virtualisation is that 3D hardware acceleration doesn't work, preventing you from enjoying your hard-earnt game of Quake 3. Rumours abound that Parallels is developing it for its software, and that VMware is doing something in this area as well.

However, thanks to the Google SoC, the power of open-source itching, and the talented Andrés Lagar-Cavilla, Xen now has support for 3D acceleration as well! Check out the xen-gl web-page with screenshots, or just clone xen-gl.hg and get hacking!

Rather than getting down and dirty with foreign grant mappings, PCI pass-through and all that malarky, Andres adopted for the more pragmatic approach of packetising OpenGL using the Chromium project, and creating an x.org module to correctly position the resulting OpenGL. End result: hardware rendering in a guest domain, without requiring any extra hardware privileges. Awesome to the max!

Dr. Madhavapeddy I presume?

anil@recoil.org (Anil Madhavapeddy) — Mon, 10 Jul 2006 12:43:40 +0100

After an invigorating 120 minutes of questioning, bright lights shining in my face, and general all-round cross-examination, my two examiners Ian Wakeman and Tim Griffin decided to pass me with minor corrections for my PhD! On top of that, it miraculously got nominated for a BCS Distinguished Dissertation award, which does motivate me to really polish it up before submitting the final version. At the Volta Lounge, we are planning to graduate together next summer, so there's no rush...

I haven't quite gotten used to it yet; while booking tickets to San Francisco for Thursday I bottled out to the sales lady and put my title down as "Mr." instead of "Dr."!

Tony Hoare on bounds checking in 1980

avsm — Tue, 08 Nov 2005 19:25:22 +0000

I've been majorly focused on finishing off my PhD Thesis recently, hence the lack of updates (but check out the sharp green gradient I'm posting on the thesisometer!). While researching the history of dynamic bounds checking in languages, I found this remarkable quote from Sir Tony Hoare in his 1980 Turing Award lecture about Algol-60:

A consequence of this principle is that every occurrence of every subscript of every subscripted variable was on every occasion checked at run time against both the upper and the lower declared bounds of the array. Many years later we asked our customers whether they wished us to provide an option to switch off these checks in the interest of efficiency on production runs. Unanimously, they urged us not to - they already knew how frequently subscript errors occur on production runs where failure to detect them could be disastrous. I note with fear and horror that even in 1980, language designers and users have not learned this lesson. In any respectable branch of engineering, failure to observe such elementary precautions would have long been against the law.

Bear in mind he made this statement in 1980, and nothing has really changed in the intervening 25 years as the Internet gets overrun by viruses and worms which take down hospitals and nuclear power plants.

Another amusing one is about Fortran, found on his Wikiquote page:

On October 11, 1963, my suggestion was to pass on a request of our customers to relax the ALGOL 60 rule of compulsory declaration of variable names and adopt some reasonable default convention such as that of FORTRAN. [...] The story of the Mariner space rocket to Venus, lost because of the lack of compulsory declarations in FORTRAN, was not to be published until later."

The reality of Japanese camera-phones and codes

avsm — Fri, 16 Sep 2005 15:33:43 +0100

One of the more interesting discussions I had during Ubicomp was with Tim Kindberg and a very nice chap from France Telecom who lived in Japan for a few years. He was very familiar with the state of barcode tagging as deployed in Japan, and these points stood out:

Most barcodes in use here are QRCodes and NTT phones come pre-installed with a reader. Users can also install a UPC barcode reader. The first such use of QRCodes you see is in your passport; the entry stamp has a barcode on the sticker with your id number on it.
QRcodes are so easy to read since most camera-phones in Japan are auto-focus, in contrast to our crappy fixed-focus attempts here. Our tests with QRcode reading using the KDDI cellphones were pretty successful due to that alone.
QRCode deployment in Japan is by no means ubiquitous, as reports in some blogs suggest. You find them on some products (like, oddly enough, tissue packets) but most advertising posters are distinctly QRCode-free. I certainly never saw people clicking on them in public over around 3 weeks of wandering around the country.
The telcos here draw a clear distinction between content providers and application providers. Phones aren't quite as programmable as in the West, and so most barcodes take you to a webpage portal with various actions (such as buying the product or just linking it for future reference). Of course, phones in Japan have cheap high-bandwidth connectivity, so this works very well without the long latencies and download times that we have to put up with on our carriers. We confirmed this by playing with the phones KDDI provided at the conference.
One really interesting example was that mobile phone bills sent out to people in the post have a QRCode on them, which, when clicked, is stored in one of the phone "barcode slots" in the reader application. Users then go to a post office or bank, and can pay that bill by pressing their phone against an RFID reader (which pays the bills for all the QRCodes stored on the phone). A superb example of a pick and drop interface in the wild.

Ubicomp Day 3 roundup

avsm — Thu, 15 Sep 2005 16:01:00 +0100

Ubicomp had a pretty strong technical start in the first few days, at least compared to previous years. The last day of the conference was more HCI-centric, but here's a quick summary of what went on from a combination of notes from myself and Al. It isn't entirely complete, as we admittedly blew off the last session to go watch the September Grand Sumo Tournament (which was unbelievably fun, more on that in a later entry!)

We somehow ended up in the Sumo house during the conference...

... and found the main ring where battle commenced!

Later, we found the sumo fanbois gathered around their fave wrestlers banners

Two guys from Nokia presented a field study they conducted on the DigiDress Nokia Sensor application (pdf, homepage) for their Symbian phones. The idea is fairly standard as these social networking applications seem to go: it Bluetooth scans for nearby devices, checks if they are running the Sensor software, and exchanges various bits of information (e.g. photograph and vital statistics).

It does stand out for being the most slickly implemented Symbian application I've ever seen, with perfect animations and feedback during the Bluetooth work. I noticed in the paper acknowledgements that they used some Nokia middleware for the networking part of the application; something that would be great to see released publically.

There weren't really many "take-home" points that I got out of the presentation however; they ran a closed test with Nokia employees, and the paper is full of various statistics and quotes, but nothing really jumped out at me. To summarize: 681 users, 80% were Finnish, and they were comfortable giving personal information out since all were from Nokia but probably wouldn't in the big bad world. I downloaded and ran the application during the presentation, and kept it active for the rest of the conference, but failed to find a single buddy who was also running it. Al mentioned that he had tried it a few months ago in Cambridge with similar lacking results.
The PlaceLab presentations continued with Control, Deception and Communication: Evaluating the Deployment of a Location-enhanced Messaging Service (pdf). They evaluated a messaging system which used geocoding using GSM cell information (so you tell people where you are, and where you might be). As Al observed, iChat's away message but used for location.

The focus of this paper was on location privacy, and they were trying to figure out what it takes for it to be useful, how to promote personal privacy, and whether or not plausible deniability was possible (an interesting question when using normal IM systems, which tend to leap online when you open your laptop, making it hard to deny that you are online, with simultaneous logins making the whole thing more complex).

I think I'm quite sympathetic to this work (despite the rather thin slicing from the Pervasive 2005 Reno paper) as they laid out some decent design conclusions that I agree with: (i) preserve plausible deniability in the interface; (ii) lightweight messaging (e.g. predefined templates intelligently popping up, much like the predictive texting) could be a killer application; and (iii) automatic message exchange should not be a design priority, as people prefer to control the degree of privacy and precision in reporting it rather than letting an electronic agent control that for them.
Next up came the Microsoft Research work on AURA: scanning objects in the wild (pdf) which uses a handheld scanner to read in physical hyperlinks such as UPC barcodes, and upload them electronically. The idea is that this can make applications such as shopping much easier as you simply "bookmark" the physical objects and collect them later, or even just go home and find the best price online.

"Sounds cool!" I hear you cry, but amazingly (for me anyway), this work didn't cite, or even mention, the work HP did on Cooltown which did exactly this with handheld devices in a closed environment. This being a research conference, I would expect that people look up related work and try to build on previous work. Anyway, even after reading their conclusions, I was left fairly unimpressed (although the study itself seemed to have been done fine). For example: end-to-end support for applications says that users wanted AURA to be integrated better in "certain" applications; I assume this means that if I scan a book in, I'd like it to be uploaded to Amazon to my wishlist. Or, integration with current practices leaves me a bit bemused; isn't it obvious that a hyperlinking system needs to fit into the way users already operate? Some points are very western-specific; robust offline usage is of less relevance in Japan where a reliable 3G network exists that alleviates the need for a lot of state on the device itself.

I think work of this nature is in real danger of just being left behind in the dust, as companies such as Delicious Monster are rapidly creating very user-friendly, popular software that just does what systems like this are trying to research. Still, the people working on it are the very smart folks at MSR, so I'll definitely be keeping an eye on it in the future.
Some quickies: the Abaris autism system (pdf) was a very solid piece of work on integrating in computer support to therapists working with autistic kids. They had some interesting observations on the usability of alternate input methods (such as voice) when in a real environment. The (very fun) folks from Glasgow talked about Treasure: picking pockets on the lawn (pdf), which is a mobile multi-user game which uses 802.11b dead-zones as part of the game tactics. I must admit this left me a bit confused as to the point of it all; why not just simulate the dead-spots in the game, and restrict yourself to depending on the environment? Its not as if users can magically perceive the dead-spots, so they can't really build tactics based upon where they are anyway, except as a primitive memory game mechanism. IBM came along to discuss frameless displays (pdf) and presented some very early design guidelines for frameless applications (i.e. set background to transparent). In living for the global city (pdf), the People and Places Lab at Intel followed three people (from London, Los Angeles and Tokyo) and tried to work out what they carried. Pointless ethnography at its height?
Finally, the best for last... some Danish researchers presented the ActiveTheatre, a Collaborative, Event-based Capture and Access System for the Operating Theatre (pdf). Capture and Access (C&A) is jargon for a system which tries to remember everything around you and your environment and make it available for future retrieval; an electronic indexed memory of your experiences. They observe that C&A systems are particularly useful for doctors in the operating theatre since they rely heavily on items like the patient's paper record, medical images such as X-rays, and even instruction manuals. They conducted real field testing with real doctors performing operations, and experimented with input mechanisms such as speech. Very impressive work indeed!

Ubicomp 2005, Days 1 and 2

avsm — Tue, 13 Sep 2005 07:37:00 +0100

I'm in Japan at Ubicomp 2005 at the moment, and have been taking notes along with Al using Subethaedit over the dodgy WiFi network here. The conference has been better than the previous ones I've been, most likely since its in Japan where we can sample some of the more exotic technology in use here. Here's some of the more interesting things we've seen in the first couple of days...

In Japan one drinks...

... eats ...

.. and enjoys the view...

Ubiquitous Networking:

Some guys from NTT presented the CarpetLAN (pdf), which is a networking system embedded in normal carpet tiles. They got 10Mb/s out of it with pretty small cells, but users need to carry a small transceiver (in their demo, it was hooked up to an iPaq). They reckon they can scale it up to 65000 tiles, so it could be a good way of tracking people in large crowded places like conference centres (since multiple user s can stand on the same tile and still get reliable connectivity). The videos were pretty cool, but don't appear to be online unfortunately.
Next up, u-Textures (pdf, homepage) talked about creating self-organising displays by having a bunch of movable LCD touchscreens which users could detach and place in different configurations for various applications. For example, a media shelf would be arranged with one screen sticking out, and another vertical on the wall. Users can then place a CD on the shelf, which would be detected via RFID, giving users the option of playing it. An alternative application is sitting around a table and organising them in the circular fashion, letting tourists decide where to go by controlling a central map interface.

This was a classic "Ubiwhat?" in my book; I just couldn't understand the point of it. From a user-centric point of view, I find it hard to believe that re-arranging screens physically is a practically useful interface. Also, each application requires a specific arrangement, and it wasn't clear to me how users could figure out the arrangement required to actually run the application in the first place.
I actually missed this talk while chatting outside, but the paper looks pretty cool. They describe (pdf) a declarative language for expressing a user interface, and a compiler which applies a cost function (e.g. number of clicks to do something) to automatically generate an optimal user interface for a given device. The device could be a PC or a little embedded device with a limited set of inputs and outputs. Only downside is of course the compilation time for the interface on a small device like a PDA (they quote 40s for their example on an iPaq), but of course it only needs to be done once and then rendering it is very fast.
The "most likely to be a joke paper", on how to analyse dietary habits using the chewing sounds that people make (pdf). The presentation was actually pretty good and fun; they motivate their work by noting that 1 billion people in the world are overweight, and that a solution to help people self-regulate is important. It could be done by keeping an eye on the type and amount of food you eat.

They place environment audio sensors around, and depending on the number of times you chew your food, get some pretty good results for some types of food (e.g. potato crisps), and not so good for others (yoghurt!). Still, I think health and fitness is an area that could be extremely well served by Ubicomp, as its something which is easily measurable, and the results easily quantifiable (unlike the social networking stuff which seems to be so loved here).
A super idea from Georgia Tech on designing Capture-resistant environments for camera-phones (pdf). Its obviously a hot topic, with a lot of areas such as swimming pools and gyms banning camera phones. This paper exploits an effect known retro-reflection in order to detect the thin film between the lens and CCD/CMOS on most camera-phones. They shine a cheap Sony Handycam infrared light and then detect the reflection. There is then a large projector in the environment which shines a beam of infrared light back at the camera to overload the CCD and ruin the picture being taken. More pictures and videos available on their homepage, definitely worth a look. Unfortunately, it only works on current camera-phones, and not with analogue cameras for example, but still a great bit of work.

Location Systems: This set of papers could definitely be summarised as the "Placelabathon", with Intel Research presenting loads of papers related to their PlaceLab project being run out of Seattle.

PlaceLab works by creating a global database of GPS and WiFi location traces. So far, they've talked about how self-mapping locations using 802.11(pdf) is practical (they got people to walk around with laptops for 3 days and tracked where they were and what they could see). Also, the prolific Jeffrey Hightower talked about a system for mapping meaningful names to co-ordinates and vice-versa (pdf); a GeoDNS of sorts.
Ali Tse and I got our turn in the spotlight to talk about the measurement work we did using the Active Bat location system to take some really accurate Bluetooth measurements (pdf). We took turns talking and I had a lot of fun, as it gave me a chance to rant about Bluetooth suckiness! Some good questions as a lot of people came up to chat afterwards to either argue or agree with our assertions that Bluetooth location is a waste of time given current hardware limitations.
The fine folks at Bristol are trying to overcome some of the limitations of the Active Bat system by creating a self-calibrating ultrasonic location system (pdf), with relatively cheap tags that cost about $100 each. The system had pretty good results (with 2.5cm-5cm errors on the calibration), and the measurement methodology was a bit suspect, but overall I think the system has a lot of promise. I think its becoming clear that the main thing that location systems need is not superb accuracy, but rather dependability and consistency over time or it loses user confidence. The number of Bat users in our lab can be counted on the fingers of one hand now, unfortunately.

Cool Japanese Stuff: There were a bunch of demos as well as papers, and they were of really good quality this year as real companies such as NTT and KDDI showed up with their funky Japanese gadgets. There were some real stinkers as well (such as IBM's bizarre PDA remote control display), but once we battled past the language barrier, these were some good ones:

The NTT Cmode vending machine (cmode) was awesome! Its a normal vending machine (which, in Japan, means it sells everything you need to survive), but you can pay by your mobile phone. It has infrared, RFID and a camera interface. A load of the phones around here have RFID built in, and this was the best way to use the vending machine (just press your phone against the machine and out pops your Coke). The camera interface works by your phone downloading a one-time QRCode to your phone display, which you then wave in front of the camera to get your drink. It was a bit more clumsy than the RFID one, but you could of course print out the barcode to paper (enabling special offers to be handed out nice and easily).

The Cmode vending machine...
... and the actual interfaces
Some of the autofocus Japanese phones to read QRcodes
The slightly bizarre but very effective Suica Direction Finding Pole was another great demo. Suica is an RFID pre-paid card, similar to Octopus in Hong Kong or Oyster in London. In this demo, you wave your card on a location on a map (e.g. in the subway), and the system remembers where you want to go. Next, if you see a "direction finding pole" (believe me, you wont miss it), wave your card at the pole, and the flexible top of the pole will swing and point in the direction you should walk (it rotates around 360 degrees in a really funny way). There's also an LCD screen for more information. After experiencing the Japanese subway system without being able to read Kanji, a system like this being deployed more widely would be fantastic for tourists anywhere (consider that you could select the location you want at your hotel at a tourist guide).
Some quickies: Fujitsu are demoing a contact-less palm/vein authentication system which worked flawlessly for us (link). IBM, as mentioned previously, had a pointless PDA demo which acts as an input for a public display. The interface sucked, the response was slow, and the demo failed. Finally, there was a cool motorcycle heads-up display designed for pit crew and spectators to see stats and track information. Although a bit pointless right now, the guy said he wanted to extend it to overlay information over the heads of motorcycles as they drove by (tracking them by GPS), which would be cool!

The IBM pointless PDA display
Awesome Seico direction finder
Ali vroooms with the motorcycle headset

Return of the Thesisometer

avsm — Tue, 21 Jun 2005 16:30:59 +0100

Summer is upon us, the sun is shining, the birds are singing, and ... the thesisometer has returned, courtesy of Andy Warfield! It's going to be a long summer ...

We make the New York Times!

avsm — Thu, 07 Oct 2004 15:54:14 +0100

A very nice reporter from the New York Times showed up at Intel Research Cambridge last week, and he just published an article titled "Connecting Paper and Online Worlds by Cellphone Camera" about our work on SpotCodes. We're still trying to find a paper copy of the article as the NYTimes is not available in Cambridge unfortunately.

This is a much more serious article than a previous one by Wired titled "From the Prawn of Time" (which we shared with the extremely cool Prawn Sandwich Clock demonstrated at Notcon). Incidentally, on the academic front, the ACM MC²R journal just accepted some of our work on using visual tags to assist in Bluetooth service discovery. Things are certainly moving!