The old Star Trek episodes are wonderful morality tales. And hey, it’s not their fault that special effects in the late 60’s weren’t very advanced, and the effects budget per episode was apparently about $50.

I smile at the cheesy effects and appreciate the storyline, but I can’t help but enumerate a few effects things that particularly tickle me as a computer scientist.  I’m blogging these so that someone blogging in 2020 can smile at my comments, and all while giving the original Star Trek the largest possible credit for couching great stories in something that the 60’s thought the future would look like (remember, the final Star Trek episodes were taped a full 2 years before the moon landing):

1. The analog dials on the ship’s computers
2. Computers with AI-complete speech recognition, but synthesized voices that sound terrible
3. Computers that actually emit smoke when they fail
4. Audio communications that fail as analog radio would (analog static and high-Q artifacts, not digital dropouts)
5. Video that fails by getting snow or loss of analog sync (I still can’t believe people used to do all this with analog signals.  That is some studly engineering.)
6. “Video” displays that are clearly posters
7. Mechanical switches for controlling electronic devices, including the transporter and the “computer”
8. Mechanical printer noises (ie, dot matrix and band printers)
9. The presence of physical books rather than ebooksat various official proceedings
10. Ship’s avionics that fail by bursting into flame, yet are apparently repairable at the board or component level

Note that a great many of these effects were actually accurate when viewed from the standpoint of the first computers in spaceflight—see Computers In Spaceflight: The NASA Experience for a most excellent overview and retrospective.

• Peter Bodik’s paper on “Fingerprinting the Datacenter” (joint work with Moises Goldszmidt at Microsoft Research Silicon Valley and Dawn Woodard at Cornell) was accepted to EuroSys 2010, where I’ll also be giving a tutorial on Web 2.0 applications;
• Wei Xu presented an online version of his work on data mining of console logs (joint with Ling Huang at Intel Research Berkeley) at ICDM 2009 last month;
• Dr. Archana Ganapathi filed her PhD dissertation (yay!!) and just had a paper accepted to the Self-Managing Database Systems workshop (SMDB 2010) on statistics-driven workload modeling for cloud jobs like Hadoop (joint work with Yanpei Chen)
• The RAD Lab will be featured in the VMware GoVirtual webzine later this month, stay tuned!

…and of course we are planning submissions to SOCC and WebApps as well.  See the students’ pages or my project pages for more details!

These were quickly dashed. The Amazon VoD UI embedded in TiVo is unresponsive, clumsy, cumbersome and unpleasant—everything the TiVo UI itself is not. When one chooses “Video on Demand” from the TiVo menu, one arrives at a menu screen featuring something like 8 different VoD vendors whose delivery medium is TiVo. Never mind the fact that as a movie watcher, I could give a flying fork whether “Jaman Movies and Shorts” or “Disney Video” or “Blockbuster” or “Amazon Video on Demand” supplies the title I want to watch; as a customer, I’m focused on content, not content vendors.

Yet amazingly, not only is the TiVo UI organized by something as meaningless as vendor, but each of the vendor submenus takes you to a different, yet uniformly awful, user interface. Here are some misfeatures common to all of them:

• Each button press on the TiVo remote has a UI response time between 700 and 1500ms. Yes, that’s right, it can take more than a second to get any visual or auditory feedback that your button-press actually did anything. This is far above the established thresholds for perceptual causality (~100ms). In contrast, when using the “native” TiVo UI, it feels snappy and responsive.
• Each VoD source has a different menu-driven UI for search, “top titles”, browse by genre, etc. The user has a simple goal: “I want to find (or browse) movies.” Yet each submenu has a different structure, for no defensible reason. As a user, what do I care which of the vendors is providing the content?
• The GUI is not only entirely textual, it is designed in such a way that less than 50% of the already-scarce screen real estate is actually devoted to browsing. The rest is devoted to TiVo templated elements and blather from the VoD vendor about how thrilled I should be that I had the wisdom to select them.

The UI is, in short, astonishingly bad.  It is all the more frustrating because this truly awful UI, which makes me want to hurl the remote at the screen, emanates from the same device that gave us the effortlessly superior TiVo UI.

What happened?

Bottom line: when it comes to video on demand, TiVo has provided an unresponsive, inconsistent, confusing and non-thought-out GUI in which various competing vendors of VoD media fight for your eyeballs with unattractive, unintuitive, arbitrarily-different GUIs that violate every basic GUI tenet, including those to which the original TiVo GUI hews so faithfully and well.  In contrast, AppleTV, for all its bashing, has a nicely-thought-out, consistent, aesthetically appealing GUI that does the one thing you want to do: find the damn movie easily, and start watching it as soon as possible. The search interface is fast and responsive; the user experience is,well, Apple; and you can usually start watching movies within 1-2 minutes of clicking “Buy”. (The last time I used Amazon VoD on TiVo to watch a 30-minute TV sitcom episode, I had to wait 20 minutes before I could start watching, even though my TiVo enjoys access to exactly the same broadband network as my AppleTV. How is 20 minutes “video on demand”? I could have made a trip to the local video store and been back in less time than that.)

Shame on you, TiVo.  DVR hardware is commodity; what had set you apart was your UI. As you continue to add vendors to your hideous VoD user experience, you will start running out of feet in which to shoot yourselves.

But hey, with disk space being free now, if I was graduating now I would also “file” a copy of the VM images used to format my thesis and run the experiments. Some of my students are doing cloud computing research so some of their VM’s are already being stored as Amazon AMI’s, but why not snapshot a VM image of their laptop as well? We’d be one step closer to truly reproducible results in CS research.

A1. Always teach. Always learn. Think for yourself. Don’t just sit there. Take Gandhi’s advice and become the change you wish to see in the world.

Q2. What two delusions would you most like to disabuse early-career grad students of?

A2.  (1) The belief that you are a good technical writer. Technical writing is more a craft than an art, and your first draft usually won’t be that good. (2) The belief that people care to read about the details of what you did. They only care about your game-changing advance you’ve made. The details are there to convince them that you’ve worked it through.

Q3. What programming language should be the first one we teach non-CS students?

A3. If the goal is understanding “how to think computationally” (to steal and pervert a phrase from Jeannette Wing), I’d say BASIC. It forces you to learn to think straight and provides “baby step” introductions to fundamental concepts like conditionals, iteration, subroutines, and variables.

Q4. What about CS students?

A4. Here the goal is exploring the Big Ideas in CS—abstraction, higher-order programming, modularity, complex data structures. So I’d say Scheme. It can be learned in an hour and elegantly expresses all of these.

• Start an autotuning DB for use by SEJITS as well as manual use. Challenge is to determine a schema for this info that could be used both for human queries and machine queries (eg via XMLRPC). Each time an autotuning parameter set is determined, add it to the DB.
• Use Archana’s and Kristal’s KCCA algorithms as as test case for “frictionless”.  They are sparse-matrix eigenvalue solver problems.
• SEJITS: take Andrew Ng et al’s paper on mapping a variety of SML algorithms to “summation form” for GPU execution, and apply SEJITS to those computations.
• SEJITS: look at LAWN 223 (Cholesky factorization on GPU) and encapsulate it in a specializer.

The good (in order of coolest to just nifty):

• PDF files look really, really good.  I’ve tried viewing technical papers (PDF generated from LaTeX or Word), PDF e-books purchased online, music printed from Finale, and sheet music and books scanned  to PDF bitmaps.  They rendered quickly and look awesome.  This alone justifies the purchase for me.  Indeed, with sheet music, there’s now the possibility of downloading MP3 files of, say, Dvorak’s New World Symphony and following along in the score as I listen…sweeet.
• The auto-rotate to landscape mode is nice, though for PDF files it doesn’t seem to always get the page breaks right if you rotate the file out of its “natural” orientation.
• The responsiveness and screen refresh feels somewhat faster than the K1, but not stunningly so, and I didn’t really have complaints about the K1 anyway.
• The industrial design is considerably spiffier. If not for the Amazon logo, you might mistake it for an Apple product. That’s not a knock, it’s a compliment. Lesser artists borrow, great artists steal.

The bad (in order of most annoying to least annoying):

• No touch screen. Not surprising but it would have been REALLY cool. I almost purchased an iRex iLiad Book Edition instead for this feature.
• Can’t annotate PDF files (as far as I can tell) by attaching notes/highlights to them the way you can with ebooks.
• The UI still doesn’t let you organize your documents and books into folders. Since I plan to use the PDF feature to read technical papers, I’ll be at 100 documents within a week. It’s going to get unmanageable fast.
• The charger/USB port is micro-USB (not to be confused with mini-USB). Not only can’t I use my K1 charger, I can’t even use my existing USB cables.  (It comes with a combination USB/charger cable, but it’s just one more adapter to carry.)
• The UI for selecting “links” in an ebook is, in my opinion, worse.  The K1 had a separate scroll wheel with a little metallic-looking “slug” you used to select links.  The DX (and I assume the Kindle 2) try to repaint an on-screen pointing-hand cursor (browser-like), but the screen’s refresh isn’t quite fast enough to keep up, so there’s a lot of ghosting and mispositioning.
• The power switch is a slider, so you have to hold it for 4 seconds to turn the device off.  The K1 had a mechanical 2-position switch that I thought was easier to operate. Similarly, turning wireless on and off now requires a menu selection; on the K1 it had its own switch.
• Unlike the K1, it doesn’t come with a protective case of any kind. I know this is to stimulate an aftermarket, but still.

The executive summary:

• I’m really glad I got this. The PDF reader alone is the killer feature. (WIthout it, I wouldn’t have cared about having a larger screen.) I’ll save hundreds of pages of printing a year just by reading technical papers on here.  Magazines like CACM and IEEE Computer are now PDF format, so I’ll be able to stop receiving paper copies of those too.
• This doesn’t replace my K1. The DX is clearly a two-handed device, like one of those leather portfolios that suits use to carry around professional documents. I’ll use it to read technical papers and other things where the large screen is mandatory. For everything else—including documents reformatted into a markup language from, say, HTML or Word or even latex2html—I like my K1: it’s small, I can use it one-handed (e.g. when standing on the train), it feels like a more appropriate form factor for reading in bed, etc.
• This is a killer device for textbooks, as I had suspected.

TiVo is widely and deservedly praised for its outstanding user experience. But in the video-on-demand department, it’s so pathetically inferior to AppleTV that it doesn’t appear to have been designed in the same decade. The great guys and gals who designed the original TiVO UI must have left the back door open when it comes to the VoD UI.

If I select the TiVo “video on demand” menu option from the top-level menu, I get six submenus: Amazon Video on Demand, Walt Disney Studios, Jaman Movies and Shorts, Browse Free Videos, Music Videos from Music Choice, and Home Movies By One True Media. As a user, what the hell do I choose? Why should the distributor of a piece of video be a first-order user experience choice? Wouldn’t most users be more likely to know, say, the name of the movie or its actors rather than the name of the middleman distributor that happens to control the distribution rights for that piece of video via TiVo?

Among those six choices, the submenus are all different. If I choose Amazon Video on Demand, I get choices like Top Rentals, Top Categories & Special Deals, Browse Entire Catalog, Search, and FAQ. The Walt Disney Studios menu item reveals Most Popular, New Arrivals, Browse, and About CinemaNow (whatever that is; the term doesn’t appear elsewhere in the UE). Even in menu items where Search is allowed, it appears at a different spot in every menu. None of the menus or submenus have anything other than a text-only, one-screen-at-a-time interface. It’s like using Gopher before we had the World Wide Web.

Finally, to add insult to injury, the couple of times I’ve actually purchased TV episodes from Amazon Video on Demand, I have to wait an amount of time that is a large fraction of the total playing time before I can start watching. To watch a 30-minute TV sitcom episode, I had to wait over 20 minutes before I could hit Play. Look, I understand about buffering, but AppleTV gets this right–why can’t TiVo?

How does this compare to AppleTV’s user experience? The top level choices there are “Movies” and “TV Shows”. Each choice gets me to an attractively-arranged screen of box-cover graphics with titles (vs. TiVo’s 1980’s text-only interface), or I can search either movies or TV shows. The user experience is identical for both movies and TV shows. If I happen to have rented or purchased movies or TV shows in my iTunes library, I see those as well. I can watch the trailer of any movie instantly and for free. If I choose to purchase or rent a movie or TV show, I can typically start watching within 1 to 3 minutes, in HD and digital surround if available for that movie. Amazon VoD is 1- or 2-channel sound and standard-def NTSC video.

Yes, AppleTV has a far smaller selection of content than Amazon, and the price per rental item is a bit higher. But using AppleTV is a joy, whereas using the TiVo/Amazon UE feels like I’ve been transported back to Gopher in the late 1980’s. Shame on TiVo for allowing this dreck to invade an otherwise excellently designed product. It’s frustrating that TiVo, a device that otherwise has an excellent UE and is by its nature an Internet appliance, falls so embarrassingly flat when it comes to a UE for video-on-demand that isn’t even usable, let alone compelling. Get a clue, guys–the iPod has 85%+ market share for a reason.