Hmm, another exaggeration from Jeff Atwood. He's getting good at these.
In my experience the metaphor doesn't stretch to RIAs. My start-up's main application is precisely one of those "RIA that pretends to behave like a desktop app", and thanks to Flex we've been able to get a good enough approximation. Result? Our users don't feel uncomfortable about it (although we have had the odd user who seemed to expect that everything would be instant even though it was a web-app). On the contrary, our users love it, and have commented on it positively many times.
I guess people care less about that last 1% when looking at a web application than when looking at a humanoid robot. Which means stretching the metaphor in this direction doesn't quite work.
I don't think he was making the point you think he was making. Or rather, I read it that he was in particular focusing on the visual interface side rather than the functionality and application mechanics.
I clicked on your profile and I assume you are referring to your website http://woobius.com ? I didn't try it out but I did checkout the screenshots. See, to me, your app looks like a web application. It borrows some of the visual metaphors of a desktop app, but the actual interface skin looks very much to me like a web app.
To take Jeff's metaphor, your app to me is like a C3PO. He's got all of the functionality and parts of a human (face, limbs, speech, etc.), but on the surface he looks like a robot, so it doesn't creep you out when he hobbles along with his stilted gait.
Anyhow, I thought the article was a bit long winded to make such a simple point, but for me the point resonated in the end.
Sure it's an exaggeration. The uncanny valley is a visceral biological reaction in parts of our brains that have spent millions of years become very finely tuned to identifying human beings. We have no such location for user interface. That said, I think there is still something to be said for the concept. If you are trying to mimic something, and get it 99% of the way there, that last 1% _can_ be source of frustration. (Jeff goes on to say because it _can_ be, it therefore is a bad idea, and you shouldn't do it. This is a poorly supported conclusion.)
For example, stupid cnnsi.com has a "main window" that has some things at the bottom that look, feel, and act like links. Yet I cannot middle click them to open them in a new tab. This annoys the bejesus out of me. Another example is desktop-emulating applications who don't have a fully-functional right-click menu like I'd expect.
By emulating other things, you put expectations into your user's mind about what sorts of functionality it should have. In this respect, it is analogous to the uncanny valley. The mimicry alone has created expectations that can cause frustration. But, that doesn't mean, that mimic'ing is inherently wrong or bad. It just means you have to do it right. You have to meet those user's new expectations.
So basically, Jeff's just saying, "If you look like X, you've just set all the expectations of expectations X. You'd better meet them. Oh, and BTW it's just like this other concept that got on the front page of reddit!"
I fully agree. This is why I can't stand using Meebo unless I absolutely have to. The app itself is amazing for what it does, but in no way will it replace my dedicated desktop AIM client.
Most web-apps that emulate a desktop app are still clunky and nowhere near as smooth to operate as their desktop counterpart. Meebo is one example, and all the web "desktop replacements" are another. You're always waiting for something to load, or javascript doesn't resize the window properly, windows that get dragged around don't move smoothly, etc.
But wouldn't it be awesome to drag a window generated within a webapp and toss it some place onto your desktop? (to the degree that the browser is just an extension of the OS -- kinda like a file explorer with more features)
It's these kind of intuitive things that I love. The only problem is being able to guess what is intuitive to whom. Else you'll end up with a ton of features that no one ever uses (read: waste of money).
Abstracting (too much), it's violated expectations - like broken promises, people really don't like it.
Taking a tangential step, computer automation is similar when it purports to be able to help you, but doesn't understand you well enough to do so; like MS's "clippy".
In contrast, Google's "I'm feeling lucky" button is automation that fulfills its promise (though the name emphases that luck is involved). Google can predict so well what I want, that this button is (very often) right. An awesome achievement. I added a simple command-line google lookup, to go directly to the page, and it's freaking me out how well it works:
Two years from now we'll be seeing that, or not seeing it as the case may be, in movies. Another two years after that it will be in the AAA video games with 9 figure budgets. And another two years after that you'll be able to do it from a Japanese cell phone.
I don't think the Uncanny Valley can exist with regards to user interfaces. They were created from scratch. And they've changed, and are still changing. And neither a desktop interface nor a web interface is set in stone.
See, the valley can exist because humans stay the same, at least to our brains. We don't madly change form or shape every 20 years. But user interfaces do. So how do our brains know a web interface looks too much like a desktop interface? It can only know because we know what both are "supposed" to look like.
But kids being brought up on web interfaces that look like desktops will not have a valley because to them that looks normal. There is no set "desktop" look or web interface "look." Its only what we've created. Create something different, and the valley will go away.
Depending on your client audience, user interfaces might not change enough for them to really feel too comfortable switching -- for example, if even the theme of my father's Windows installation changes, he has a lot of problems getting work done.
If you have a web app that looks identical to a Windows/OS X desktop app, but behaves oddly, it will irritate your customers in subtle but significant ways.
But isn't that an expectation problem? It doesn't behave as expected. But visually, it could have looked like anything. As long as the second thing looks exactly like the first but behaves differently, it will be irritating. But the first thing can look like anything you want and it'll never be annoying, unless it too looks like something else. Plus, someone who gets introduced to the second interface first will have that interface become their reference point.
Office already had a pretty poor interface, so maybe it's more to do with that?
I was pretty disappointed they didn't try anything different, but I guess the aim of that was to attract people who actually like Office in the first place.
Eh, there's something to be said about web applications and the Uncanny Valley, but this isn't it.
Atwood, I think, is saying that web applications shouldn't be like desktop applications because it violates the user's expectation of how a webapp should be.
Applying the Uncanny Valley to this says that it's not bad for web apps to behave like desktop apps -- it's bad for webapps to signal that they have behave like desktop apps, and then fail to live up to that expectation.
This problem affects other products, too. For example, imagine a really awesome semantic search engine called CantorSet. You can throw all sorts of questions at it and it appears to come up with the answer. "What did Bush say at the last G8 meeting?" Wow! Amazing!
But then you ask questions like, "Why did my wife leave me?" No answer! Wow, this thing sucks.
It's less about what the webapp does or doesn't do, and more about the expectations it sets and whether or not it lives up to them.
There's a big difference between web app and web page. Common web app controls (accordion, tab panels etc) allow developers to convey lots more information on screen and well thought out use of AJAX can make a complex application perform more quickly than the static equivalent.
As the complexity of web apps increases users will need to get used to the idea that they need to learn how to use each app to get the most out of it which is how it works with desktop apps. Try to compare the UI between desktop apps such as a music editor, 3DSMax and MSWord. Each uses different UI controls (sliders, knobs, text input etc). Each app is complex and has an associated learning curve. If people consider the app useful they will invest the time in learning how to use it.
I recently had to learn how to use Aftereffects. The UI and work flow were completely foreign to me - none of my existing knowledge was transferable, but I learned how to use it because I needed the features it provided.
Ultimately the market will decide. My guess is that the line between desktop and browser will continue to blur and will one day disappear.
In my experience the metaphor doesn't stretch to RIAs. My start-up's main application is precisely one of those "RIA that pretends to behave like a desktop app", and thanks to Flex we've been able to get a good enough approximation. Result? Our users don't feel uncomfortable about it (although we have had the odd user who seemed to expect that everything would be instant even though it was a web-app). On the contrary, our users love it, and have commented on it positively many times.
I guess people care less about that last 1% when looking at a web application than when looking at a humanoid robot. Which means stretching the metaphor in this direction doesn't quite work.