Google Mail’s Apple’sque feature

When I hit send on an email today in gmail’s web interface, I was greeted by the following pop-up:

Google Mail notification about missing attachment

Google Mail notification about missing attachment

This feature reminds me of what I have always appreciated about Apple’s (hardware) products – small touches that signify true polish. Nice work Google!

Buying robustness in exchange for efficiency

Over the last few decades (the last three in particular), we have seen a relentless drive for supply chain efficiency. This drive has resulted in continuous improvement and a significant reduction in cost of many of the products of our modern economy. Now, we see that same application of efficiency applied to the technology infrastructure used to build companies with the increasing penetration of cloud infrastructure. I happened to read the most recent post on Ben Horowitz’s, Revolution at the Edge, which discusses the inevitable destruction of traditional IT and its replacement with three major ideas, Cloud, Mobile, and Social.

I do not doubt that we’ll continue to see the increasing adoption of “cloud” (I’ll use it as a noun for convenience) over the next few years and the benefits that we’ll get from that adoption in terms of reduced IT costs and frustration; however, to use the terms of “revolution”, every revolution also has some innocent victims whose loss may be severely felt by society. In this case, I believe that we have significantly increased the fragility of every part of our ecosystem that has been besotted by a furious drive for efficiency.

Let me say this again: We have all chosen fragility over robustness in the choices that we have made for supply chain efficiency. Four recent examples come to mind:

1) The Thai floods that led to an overnight decrease in hard drive production capacity and a significant increase in costs
2) China’s decision to ban rare earth mineral exports
3) The loss of parts of Amazon’s data center in Virginia which resulted in several major sites like Netflix being out
4) The loss of power across most of North India, affecting over 600M people

Which one doesn’t fit the pattern?

When power was lost across North India, hundreds of thousands of generators and power inverters immediately turned on and life kept going for a majority of society. In fact, society in India is designed for loss of power, and as such is robust to power failure, at least for its basic functioning. Compare that to what would happen in the United States if the entire Eastern seaboard lost power.

India’s infrastructure is robust not by design but by accident and circumstance. Could one design a much more efficient and resilient system? Yes, of course. However, as examples #1, #2, and #3 above demonstrate, we do not even include robustness in the design consideration of our broader supply chains.

My recent area of concern has been security. I always thought that Amazon, with its AWS infrastructure, could have 10s or even hundreds of people working on security while if I colo’d my own server, I’d have one ops person running security. Clearly, Amazon can protect against a lot more. That said, once Amazon’s security gets compromised tens of thousands of companies will be compromised instantaneously since they are all relying on Amazon to have robust security. Another example of the trade off between robustness and efficiency. If each of these companies had different security infrastructure, it would be just a little bit harder to penetrate all of them at once. I’m sure several of them would be compromised much more easily as well.

And of course, one could point out the obvious – nothing in Amazon’s infrastructure suggests that you can’t implement several layers of security on top of it. However, the point is that we, as humans, don’t. Why didn’t we keep the other rare earth mines open? Why didn’t we maintain production of hard drives elsewhere in the world?

A parting thought: how screwed would Apple be if the US and China entered into broad economic or military conflict? And don’t say it can’t happen – the Proview case was a small case and yet a huge thorn in Apple’s side. We should design for failures factoring in their respective likelihoods.

A special thanks to the books Daemon and Freedom, both by Daniel Suarez. They first highlighted this design tradeoff to me.

Painting light

On recent TWiP podcasts, Alex Lindsay suggested a couple of cool product ideas:

– using a computer and one or more projectors to “paint” light onto a studio scene e.g. you could dial-in that you want a certain part of a scene (e.g. a model) lit with warmer light and the software would change the image that’s being projected with the desired light colors and shape. Of course, you’d need a lot of light out of the projectors to light any substantial set up. This idea, however, seems eminently doable. Throw in advances in LED lighting over the next couple of years, and there’s some good potential for a portable lighting configuration.

– a camera manufacturer that solely focuses on manufacturing the lenses and sensor and ships a simple body that puts the two together. In addition, it provides a slot to drop in your iPhone and makes available a simple, but comprehensive API, to control the hardware. Then, developers can develop UI’s to meet different classes of use cases e.g. HDR, time lapse, basic, etc. As they discussed on the show, the manufacturers have demonstrated a complete inability to design a good user interface in the menus or in the hardware.

As I mentioned in my earlier post, I did find the physical controls of the Minolta Maxxum 7D very compelling; the potential manufacturer would have to put in some creative thought on providing programmable physical interfaces on the barebones camera as well.

Focusing screen no help

Billo pointed out that a focusing screen would not be much help with manual focus while shooting a movie with my new Canon DSLR, since the mirror is flipped up / open while you’re shooting a movie.

Guess I’ll have to opt for an LCD viewfinder loupe instead.

Nissan GT-R

I’ve always loved driving fast cars (perhaps it has something to do with the time I spent in Michigan). I recently got the opportunity to drive the Nissan GT-R. It’s a very fast car: 0-60 in around 3.4 seconds powered by a twin turbo V6. I was half expecting (and hoping) that Nissan had done something magical and gotten rid of the usual turbo lag, but alas, that was not the case. Still something to be said for that big engine kick, supercharged or not.

Of course, once you got going, you got going! Rolling launches were also much more responsive.

A very nice touch: the center panel gives you detailed metrics from any of the car’s many sensors. You can have it display g-forces, clutch positions, temperatures, etc. Of course, while you’re busy accelerating to very high speeds on non-track surfaces, you have little mental bandwidth left to look anywhere but at the road.

After my drive, another friend told me that Nissan got the Grand Theft Auto designers to design the panel – I have not verified if that was actually the case.

All in all, neat car – would love to play with that paddle shifter next time around.

Minolta -> Canon

I finally bit the bullet and moved from being a Minolta guy to a Canon guy. I wanted to start investing more in photography gear and wanted to do it with a system that’s more common and has more options available. I tried out both Canon and Nikon cameras, and ultimately decided to go with the Canon T2i. I looked at the 7D (and longingly at the 5D Mark II), but ultimately decided to save money on the body and invest in a nice lens instead.

Took a while to get used to all the different controls on the Canon – it’s still an on-going process. I have to say, even though my Minolta DSLR is old now (~4 years, which is ancient these days), it still has far more ergonomic manual controls. Everything on the Minolta Maxxum 7D is adjustable with a physical control and the controls are laid out well. With the Canon, I have to fuss around with electronic display menus much more. But I am loving the Canon – my pictures are much sharper than before, thanks to the image stabilization in the lens, as well as higher quality optics.

I do also like the ability to record beautiful 1080p, 24fps video from the camera. I have so much respect for the focus pullers who work on TV shows and movies. I’m thinking of investing in a split-prism focusing grid for the Canon so that manual focusing in video mode will be much easier.

Before that though, the next stop is going to be: 50mm, f/1.4 lens.

Count Something

Atul Gawande, in one his recent books (maybe Better) mentioned that one should count something all the time. I recently took a taxi to the train station and the cab driver mentioned that previously, of the people heading into the T with bags, 6 or 7 out of 10 would instead come to the taxi station right next to the T stop and take a taxi instead.

As a sign of the downturn, he mentioned that he now sees only three folks out of ten take a taxi. The rest decide to just lug their big bags down the stairs and all around the T-stops.

Daniel Was Right

Daniel was right when he suggested that a daily post was quite ambitious. It indeed has been. I have been generating ideas for blog posts and writing them down in various places, but am only now starting to turn them into posts.

Podcasting Talent

As you may have noticed, I’ve been spending a lot of time with digital photography these days. On my commute to and from work, I have ample time to listen to podcasts (esp. at the 2x speed that my iPod allows). I’ve been listening to podcasts hosted by various folks, and I’ve been struck by my own preference for particular hosts.

My preference has nothing to do with the content of the show – all of the podcasts I listen to (e.g.,,,, and – have fairly good content, but some of the hosts (Chris Marquardt and Frederick Van Johnson) are a significant cut above the rest when it comes to their style.

Similarly, I’ve noticed with the MarketWatch show on NPR at 6:30 ET (in the Boston area) that Kai Ryssdal usually hosts. I find the show far more entertaining and engaging when Kai hosts it than when other folks host (while he’s out for short periods). I’ve tried to listen more closely to figure out exactly what’s the difference – cadence in speaking, intonation and variance of both tone and speed of speaking, the actual content itself – and haven’t quite been able to identify the “basis vectors” yet.

1080p HD on 1/6″ sensor

Great post today on CrunchGear about how resolution has nothing to do with image quality. Has an excellent in-depth walk-through of how light enters a lens and ultimately gets written out to memory and everything that can turn the image to garbage in-between. I vaguely knew this process; the post really goes into great depth while keeping the explanation relative easy to consume.

The motivation for the CrunchGear post was the announcement of a 1/6″ sensor from Omnivision that records 1080p.