making time

Federation and cash cows

2009-06-03T21:20:06Z

Google Wave wants to reinvent email, the original killer app, but it raises an old question: who actually made money from email? Thinking back on computing history, I can think of 4 killer apps that are in a league of their own: the window-based operating system, the office suite, email, and web search. Three of those four are industry-defining cash cows: Windows and office for Microsoft, and web search for Google. That makes email the odd one out; sure some people make money from the technology, e.g., spammers, but email does not fund an Microsoft or Google-esque empire. How can something so fundamentally important to the world reap so little profits? Well of the four killer apps, email is the only one that is federated, and couldn’t exist any other way. Painted in that sense, email is a rather heroic transfer of economic surplus from its creators to the users. Regardless of whether or not Google has some genius plan to make Waves into its second cash cow, hats off to them for taking one of the boldest moves in technology in a long time.

Lots of other things have been said about Waves, so I’m not going to revisit them. I particularly like Tim O’Reilly’s take.

A new mashup experience

2009-04-30T05:12:03Z

There’s a lot of controversy over the politics of DiggBar — over ownership of the content and control of the landing URL. However, I don’t really see it so much as a political issue as it is a technical problem. The fact is there simply isn’t a good way of overlaying 3rd party functionality into the browsing experience without resorting to a plugin install or hijacking the outbound link like the DiggBar does. I don’t really believe it is Digg’s intention to steal anybody’s “link juice”, but their actions result from a technical limitation—either they settle for an install, which would have appalling uptake, or do the nasty iframe. What’s a potential technical solution? We need a new class of browser plugins that 1) activates immediately without install 2) dispatched from the cloud, thus portable outside of a single browser installation, and 3) protected from malicious cross-site-scripting

A new class of browser plugins

What I’m really describing is really a fast, safe, and fully managed framework for accomplishing cross-site-scripting, a concept generally associated with malicious deeds but is also the underlying functionality most social toolbars are trying to accomplish.

To distinguish these from regular browser plugins, I will henceforth refer to them as XSS (cross-site-scripts):

Now how do we achieve instant-on without install (criteria #1)? I think this is the easiest of the three, as the Greasemonkey architecture already allows you to instantly enable functionality overlayed over 3rd party sites. While Greasemonkey scripts are limited to Javascript injections and lack the full power of Firefox plugins, I’d argue that most of the battle is already won with just JS. You can implement a variety of content and functionality overlays including the DiggBar, StumbleBar, and even the Meebo chat bar.

The portability criteria (#2) is based on the idea that each XSS is a piece of enhanced functionality sanctioned by a particular user, and thus should be attached to that user wherever he goes, NOT attached to the browser installation. By electing to allow the DiggBar to appear over other web sites, I’m really setting a browsing preference for my Digg user account, not a preference within the browser. When I logon to Digg from a friend’s or a public computer, that functionality should follow me, and the opposite should be true when a friend borrows my computer and signs on with his credentials. That means we need a web service to manage a user’s libary of XSS and dispatch/activate them all at once whenever a user logs in from a different computer. If I want a Meebo bar, a Digg button in the context menu, and a Gmail notifier, I shouldn’t need through the motion of signing into all three to enable that functionality!

Ensuring safety is probably the hardest one, and is the one thing that can make this whole idea buckle. I can suggest a few precautions, but I most certainly wouldn’t catch all the corner cases. Functionality of XSS should be limited to approximately what Javascript can do. That means XSS should live in a runtime that cannot access the file system (but a sandboxed DB can come in handy ala Google Gears) nor touch the local OS in any way. To prevent theft of sensitive data, all XSS should be turned off by default for https pages.

User choice

Another benefit of this XSS model is that it shifts the choice back to the user in terms of what kind of enhanced functionality should be activated. I really like the SnapShots popups, and I have deployed them on this site as the webmaster. But it what if I, as a visitor wants the same functionality on other sites? Sure I can go install the Firefox plugin, or worse the Explorer plugin in .exe form, but I belive this will never become standard behavior for non-techies. To drive uptake rates, we should start pushing browser functionality and behavior preferences into the cloud as well. With something like the Meebo chat bar , I find it rather backwards that site owners have the burden and choice of implementing what is essentially an overlay interface, when it fact, it makes much more sense as the users’ choice. The browsing experience is increasingly comprised of multiple content & service providers cohabiting a single tab/frame. Thus far, this kind of mashup experience is initiated and implemented by the site owner. With a plugin framework that has a real shot of getting installed by the masses, we can have a real user-selected mashup experience.

Overlapping shares

2009-04-29T05:30:55Z

Yes, there’s a deluge of data from social media, and people are clamoring for proper filtering mechanisms. Before you even start a conversation about filters, there’s a quick-win to reduce the noise: collapsing identical shares from multiple sources. As I comb through my daily news sources, I’m starting to bump into the same news entry or awesome-must-see-video! from different people or blogs. Now, if all those shares happen to be on Friendfeed, then it’d collapse all of them into a single line item and just credit that share to multiple sources. This needs to happen everywhere for any post that is essentially an redirect (in other words, a link to the content without any real additional content or commentary). This is especially necessary on Twitter where I hear about the same viral link multiple times from different people, often days apart!

There’s two ways to fix this. The temporary patch is to have smarter aggregators that can spot non-unique or “redirect” content and collapse them, but that’s hard work and isn’t 100% accurate. The squeeky-clean solution is to develop formal, machine-readable conventions for labeling this type of content, but requires essentially adding meta-data baggage to things as simple as tweets =(

A departure from freemium

2009-04-29T05:31:37Z

Quake Live: say the technology matures, and scales. How hard will it be to fork the game and bring back the old mods? And if so, could they create a marketplace for modders and collect reve-share ala iPhone App store? Best part about this is that the Quake 3 mods already exist, the experiment is in porting over and launching a market where real-money is spent on these mods. The one other site doing something similar, but with flash games is Whirled.

A browser feature request

2009-04-29T05:31:53Z

I have a simple feature request for browsers: a visual indicator for which tabs are playing music

Between Hype Machine, The Next Big Sound, and The Sixty One, I use my browser increasingly more as a music player; and managing music vs content tabs is a rather unrefined art. Too bad though, for as far as I know, this is technically quite challenging for even a FF plugin to do as it requires tapping into the Flash plugin’s run-time environment. Another reason why the browser UI needs an overhaul and become as robust as that of an OS.

The innovation is in social messaging

2009-04-29T05:31:58Z

I’m totally digging the concept of a MySpace email product. I’ve expressed before the importance of bridging social networks with day-to-day communication, and we’ve heard before that Mark Zuckerberg identies Facebook as primarily a communication tool. What traditional social network provides is a form of what I’d call “passive” communication—users generating feeds of information and consuming it in a distributed and asynchronous fashion. Email, on the other hand, is active communication—with people usually addressing one or few individuals with a direct message.

But why are they separate? Its clear that either form can benefit immensely from being connected to a social graph! Email can be prioritized by the graph, or it can generate a prioritization based frequency and reciprocity of contact. New forms of direct messaging that used to clog up the inbox can be categorized and presented in new UI regimes: for example event invitations can be pulled out of the inbox and into a calendar layer, social game envents (like being bitten by yet another vampire) can be grouped into its own river of minor news, messages can be easily visualized by contact. Here’s another idea: imagine adding pictures and videos to email without the the dreaded “attachment” process—just pull content from the network’s internal libraries.

Social networking will be the best thing that ever happened to email (since webmail), and vice versa.

Futurists

2009-04-29T05:32:02Z

Technological prediction is a tough art:

Worse yet, you look like a fool when you get it wrong! But every techie and VC is tasked with attempting exactly the kind of prediction above.

Trading down performance

2009-01-10T21:10:06Z

When Blu-Ray beat HD-DVD, it seemed like an historic moment of validation for Sony. They had won the next format war, and can look forward to the next half a decade owning the standard distribution format for video. Not to mention, those damn Betamax jokes can finally go away.

Too bad the next standard distribution format for is actually choppy, grainy, SD video streamed over YouTube, Hulu, Net-Flix, and others alike.

Perhaps its a matter of diminishing returns, but with video, the masses have cast their vote for convenience over quality. It happened with processors too: around the end of the Pentium age in 2003, Intel recognized that pumping out more powerful CPUs isn’t going to help you check your email or surf the web any further — what accounts for most of consumer usage anyway. Instead, people needed to use laptops on-the-go, so power consumption and embedded connectivity were the real problems that needed to be addressed. This trading down of performance for mobility is the winning philosphy behind the original Centrino and Intel’s continued control of the laptop space today.

On another note: are HDTVs a dud then? I don’t think so. However the primary benefit of HD is not going to be superior picture quality, but the ability to render a robust user interface to pave the way to more interactive living room experiences. Some recent announcements at CES corroboate that…though I do think the real winner will be a set-top box that can be easily and cheaply attached to existing, not so high-end HDTVs.

The search market share game

2009-01-07T21:48:09Z

I’ve noticed that Microsoft have begun inking deals with PC OEMs to include Live Search as the default search engine. Besides the obvious comparisons to the Netscape vs Explorer bundling conflict, there’s one other interesting facet here:

There isn’t an obvious way to “buy” market share when it comes to search. Unlike direct monetization where lowering prices will attract greater volume, ad-monetized businesses don’t really have a pricing lever one could pull to influence market share.

Microsoft has tried a cash back campaign which should prove to be an embarrassing failure any day now. You simply can’t win everyday consumers over by paying them pennies at a time. On the other hand the OEM path seems like the right way to go, for now. After all default-bundling is MS’s sweet spot, but it does raise the question: what are ultimately the true drivers of market share in search?

Does anyone really know the answer to how Google became #1 in search? I have strong doubts about the most common explanation: that superior search quality is the king-maker. I highly doubt that unsophisticated internet users switched to Google purely for notably superior search results. But what else could it be? A function of branding and momentum? but didn’t Yahoo and Lycos have plenty of that? Was it the simplicity of the Google brand and front page?

Update

This iBanker prediction caught my eye:

Prediction: Increased Competition For Search Distribution Via Toolbars, OEM Deals, and Partnerships

Winning over toolbars, OEM deals, and partnerships are all examples of securing default positioning and is a hot battleground between every search engine. Again, I suspect most users carelessly follow the whims of their default search box and are rather apathetic to the quality of search results between the big three.

The role of the browser

2008-12-28T07:00:50Z

With Google Chrome, it is ample clear that browsers will evolve into application environments. Its first design give good hints as to where this is all going. For example, the inversion of the tab bar and navigation controls is not merely a cosmetic preference, but a reversal of the containment hierarchy: where as before, navigation was a universal control which manipulated tabs, now tabs are a universal control (task management) which manipulates work flows (some of which are browser like, others not so much). The bigger picture here is that the browser is taking on some of the responsibilities of the OS GUI, specifically, managing applications. As more user activities such as music playing, document editing move inside the browser, the existing browser interface will quickly become unfit for the job.

From tabs to tasks, the evolution of web activities

Computers have made us all into multi-taskers. The original innovation around tabs is conceptually similar to the collapsing of task buttons introduced in Windows XP: group a bundle of similar tasks under a single container. And similar they were. In the old web, the everything you did was more or less a browsing sequence of documents connected via hyperlinks. However in the past few years, web activities have evolved significantly away from sequential document browsing. How would you classify the activity of having Pandora in the background? Or editing a new blog post? These activities would originally reside in WinAmp and MS Word and have little resemblance to document browsing, and thus should have little resemblance to a browser interface.

Classically offline activities are now shifting into the browser window. They are diverse in 3 major dimensions: (1) their multiplicity, (2) degree of engagement, and (3) control schema compared to traditional document browsing.

Pandora, for example, is a single-instance (1) activity that usually occurs in the background (2) and almost never uses the standard back-forward-reload-locationbar browser widgets (3). Blog editing, other otherhand, is likely a single-instance foreground activity with again, totally different controls. This necessitates different task management interfaces beyond a horizontal tab list. Does it make sense that Pandora occupies a whole tab, like its a document waiting to be read (2)? No, I think many people would prefer to ”faviconize” it. Single-instance apps, such as email or RSS readers make more sense as master tabs or container groups, not as yet-another-tab in a sea of open documents (1). Finally, the standard browser controls don’t always make sense for highly specialized, non-document and non-browsing applications (3). Take the email example: I rarely ever use the back-foward button, and certainly never the location bar (unless I navigate away). Back-foward interfaces make no sense in email: you’re navigating through a hierarchy (inbox->label->conversation->message) as opposed to a sequence.

Borrowing from prior work

So what is the correct GUI behavior for a Pandora or word processing task? Just look to desktop applications! The Pandora air app sits inconspicuously in the system tray, and when active, doesn’t feature the standard browser navigation controls. Popular blog editing tools are single-instance and sit outside of the browser, and again, feature no navigation controls.

The app management problem has already been solved before, just look to the behavior of desktop applications and how they interact with the OS’s provided task management interfaces (e.g. taskbar + system tray, dock + expose)

Browser developers should be keen to borrow these interfaces and extend such access to web apps.

The road forward

I think ultimately, this is about web apps catching up to desktop apps in richness of their GUIs. You can have the sexiest possible GUI for your webapp, but it’s still constrained within the window. Can a web app pop up balloon notifications? Run as a minimized background process? Register contexual actions on desktop icons e.g., “Upload to Box.net”? The true role of the browser will not simply to “display and view webpages”, but to bridge every available aspect of the GUI with a web application.