Remembering the Dynabook

My recent post on netbooks reminded me of Alan Kay's original 1972 Dynabook concept (pdf).

This is a companion discussion topic for the original blog entry at:

Bravo Jeff, 1000 words to tell us that netbooks are useful. My pet stuffed Tigger could have told me that and I would not have wasted 5 minutes reading it. Get back to blogging about interesting and useful things before the crap outweighs the good stuff on your blog.

Was Alan Kay’s vision valid and correct?

What is the time line on the better chipsets?

You’re getting way to caught up in the hardware, the DynaBook was as much about the software as it was about the hardware.

I agree, but the hardware problem is easier to solve. The software Alan envisioned may not even happen in our lifetimes (IMO). Some interesting links on the DynaBook software here:

instead they will put lighter batterys in order to reduce the weight and price.

Quite possible – there’s a huge amount of room for them to improve, as the old 945 chipset is not even remotely power efficient relative to the Atom. It’s accountable for 80% of the idle power draw!

Get back to blogging about interesting and useful things before the crap outweighs the good stuff on your blog.

You should totally demand a refund!

Interesting timing! At the Computer History Museum tonight (5 Nov 2008), they’re having a talk called The 40th Anniversary of the Dynabook where Alan Kay, Chuck Thacker, and others will be speaking.

1 Like

I cringe every time I use a browser for many reasons. The browser people had a chance to make a more integrated UI and functionality
Is there an expansion on why this would be a good thing? Seems to me that you don’t particularly want UI and functionality to be integrated on a flexible medium like the web.

And is it me, or is it hilarious that, at a time when the whole world is watching an election that will be studied for generations to come, I’m reading about the Dynabook.

Well, I thank God for the notebok, I would be able to take a pc the the bathroom!

That is usually the way with revolutionary concepts, they are years ahed of relity

The revolution will come when they find a way to make a touch screen that can be bended, imagine the possibilities. For now the laptop like devices are the imminent future.

Revolutions in CHI are all hardware driven. Unfortunately software just try to tag along.

You’re getting way to caught up in the hardware, the DynaBook was as much about the software as it was about the hardware. It was a vision of a computer that would allow personal mastery, where every part in the software could be inspected and changed, and combined in new and interesting ways. It was a vision of a platform where users could easily define and run simulations to gain greater insight into an area of interest, meaning it was well beyond the capabilities of dead trees.

Think of discussions about creationism and evolution. The parties could implement evolutionary algorithms and test the power of natural selection and expand their understanding of that mechanism.

Windows and Mac PCs doesn’t even begin to approach this. To get a clearer idea of what Alan Kay wanted look into Squeak, Etoys and Scratch.

Apparently Kay doesn’t think much of the current status quo, where you define the status quo as OS X, Windows, or Linux. I suspect much of Kay’s objection to the web browser interface is the general passivity of browsing the web; bear in mind that Kay is an educator and originally intended Dynabooks as tools for children to create and explore with something like Logo.

I’m always amused whenever Mr Kay is quoted as a interface ‘expert’,
have a go at his Squeak Smalltalk to see a really appalling UI, perhaps the worst readily available example of how not to interface to a computer system, current browsers are easy to use by comparision !

That’s a pretty sweeping judgment on Intel’s engineering of the 945GSE chipset which, by the way, was designed for the Atom and launched in Q2 2008 so is hardly creaky old.

Typically for a software engineer you assume that the computer is primarily all about the CPU and that the chipset is just a little bit of connective tissue to the I/O that you program against. In actual fact the chipset is a highly complex chunk of functionality in its own right. The 945GSE chipset runs the MMU and memory controller, PCI bus, SATA/ATA-100 HDD ports, USB ports, audio (AC’97 and HD), 10/100 ethernet, 2D and 3D graphics and so on (check out the data sheet for the 945GSE for an idea of its complexity). Unsurprising with all that disparate functionality it’s a lot harder to extract power savings than it is with a CPU.

Much of the Atom’s power savings come from using a simple in-order execution strategy (as opposed to the much more complex out-of-order execution that all mainstream CPUs use these days). Most of the interesting choices that were made in the design of the Atom were around mitigating the effects of pipeline stalls that occur on a cache miss, which is the primary drawback of an in-order architecture. There simply isn’t the same low hanging fruit that can be picked off in a chipset to get the same power savings.

Incidentally, the Atom is a great example of why you can’t simply compare clock speeds even for CPUs that support the same instruction sets. Comparing the clock speed of an in-order CPU against an out-of-order CPU is literally like comparing apples with oranges, and that’s before you consider the effect of cache sizes.

The next generation of the Atom will include an integrated memory and graphics controllers on the same chip which will undoubtably reduce overall power consumption, but don’t look for the same orders of magnitude in power savings that was achieved with the CPU. Fundamentally the Atom and its chipset was designed to good enough and with 2 to 2.5 hours battery life being the norm for all laptops these days I think they hit the nail on the head.


The 945 chipset also brings unncessary power consumption. If you take a look at the Mini-ITX board Intel designed for the Atom, the heatsink configuration seems superficially normal: a tiny heatsink on the chipset and a larger one with a fan to cool the processor. But it’s actually the other way around: the tiny heatsink is all that’s needed to cool the Atom, while the 945 chipset requires more robust, active cooling. It seems counterintuitive that an x86 processor can sip power and run happily with the most minimal of cooling while the less complex chipset behind it is producing the lion’s share of heat and drawing the majority of the power going into the machine. One wonders how much smaller still Asus might have been able to make the Eee Box if Intel had turned its engineering genius on the chipset supporting the Atom.

It’s astonishing how power efficient the Atom is, and how UN-efficient (or perhaps a better word is typical) the supporting chipset is.

madse nails it in one: the DynaBook was as much about the software as it was about the hardware - though I would have gone further and said it was more about the software.

Being a x86 processor means very little these days, most of the current architectures emulates the x86 instructions using their own internal instructions.

A larger-format iPhone would be perfect - something a little smaller than a legal pad that you can browse the web on and take notes on.

@ Herb
Rather difficult to hold a legal pad sized device to your head to take calls though don’t you think?


Alan Kay’s comments remind me of his time in the 1970’s, when I first started learning about computers. Back then, people were saying that if we only taught kids to program, it would be nirvana.

It’s the Blank Slate meme. Negroponte’s been pushing the same thing with OLPC. (Kids were supposed to learn by themselves with it! No teachers!)

I’d be happy to have just a mundane web-browsing Netbook. I’m old fashioned and not dynamic anymore.

Easily the most pointless thing i have read today


The seeming disparity in engineering genius between processor and chipset has more to do with economics than anything. Intel manufacturers many of their chipsets in depreciated fabs (hey, they’ve gotta make something), meaning that they’re a couple process generations behind. Squeezing every last dime out of a fully-depreciated asset is smart business sense, even if it creates Odd Couple situations on Atom motherboards from time to time :wink: