Ravi Pandya

Fri, 31 Jan 2003 17:00:21 GMT

More Phil Salin: The Wealth of Kitchens

I was inspired to reread another of Phil's essays, The Wealth of Kitchens, bringing to light all the accumulated knowledge that is implicit in the artifacts and habits of everyday life. In the process, I found a site of papers by and about him, so I've updated the previous link to point there as well. Enjoy!

Thu, 30 Jan 2003 15:07:29 GMT

Freedom of Speech in Software

The discussion with Miguel about Microsoft's patents reminded me of this 1991 essay by Phil Salin. He wrote it in response to the PTO's request for comments about whether they should continue to accept patents on software inventions. Phil's cogent and impassioned plea to consider writing software as an exercise of our right of free speech is inspiring. Unfortunately, of course, it was unsuccessful, as were the many other letters that I'm sure were written at the time.

I helped him a little bit with the final editing and submission. My last memory of Phil is of him dictating some last-minute changes from his hospital bed in Stanford, where he was diagnosed with terminal liver cancer. I miss him.

Wed, 29 Jan 2003 14:49:37 GMT

I went to the crossroads (...)

I enjoyed the blogger bash with Sam Ruby at Crossroads in Bellevue last night. It was great to put faces to names, and to meet some new and interesting people. Some snippets:

Don Box teasing Miguel de Icaza about his "homage" (pronounced the French way, like "fromage") to C# and Outlook. Miguel de Icaza remaining good-natured and enthusiastic anyway, and graciously refraining from mentioning C#'s homage to Java. Mono has been making amazing progress. I remember when I first saw it I said that they might well ship before Microsoft, given the necessarily glacial release cycle of a 40MLOC OS. Miguel said they've been talking to Red Hat about including Mono, but there are some concerns about Microsoft's patents. I think there's still a good chance you'll be able to buy a shrink-wrap Linux distribution with Mono before you'll be able to buy a shrink-wrap version of Windows with .NET. (Yes, I know you get a lot more goodies in Windows .NET, but the point remains; in fact that is the point: more=later.)

I asked Miguel about Mono on OS X and he said that it runs now, but only with the interpreter. However, the next version of the jitter has a more easily retargetable code generation using a finer-grain abstract machine language where they can do instruction-level scheduling and optimization before the final translation to native code, and it takes 30% less time than the original JIT. Slick.

Don Box saying that if we're still using C#/Java style languages 5 years from now, we're fucked. Maybe I'll finally get to use Smalltalk again. It's still the most productive programming environment I've ever worked in, though Java with IDEA comes close, and even surpasses it in some ways.

Some interesting links that came out of it: (other than another 10 blogs on my blogroll :-)
Sun Labs Ace Project
UMD DateLens calendar interface
Modern C++ Design: Generic Programming and Design Patterns Applied

Sat, 18 Jan 2003 18:31:47 GMT

Molecular Electronics at HP

I went to see a great lecture this week at UW Nano by Duncan Stewart from the molecular electronics group at HP Labs. They have a great multi-disciplinary group there, with computer architecture, physical chemistry, organic synthesis, polymers, electrochemistry, materials science, and experimental & theoretical physics. They're using a cross bar architecture with Langmuir-Blodget molecular monolayer between a grid of contacts in a crossbar architecture. Not only have they built a 256-bit memory (to store "HPinvent" :-) but they have also configured it as an FPGA cell with 2-bit multiplexers on the input and output, and a lookup table in between. All this in a square micron!

One of the interesting tidbits was from Stewart's own work in trying to characterize and understand the actual operation of the device. They first tried a monolayer of Jim Heath's fancy rotaxane molecules with a movable ring system acting as a switch. The I-V curves show a nice negative differential resistance that can be used for switching and diode behavior. OK, great, it's the little ring moving along the backbone of the rotaxane. But then they tried a control substance - eicosanoic acid, "basically floor wax". The numbers were a little different, but qualitatively the behavior was the same.

So it's not the material, but probably some interface effect. But what is it? This is a real puzzle, and it's not solved yet. The best hypothesis so far is temperature dependent tunneling effects at the metal-organic interface, pretty much independent of the particular organic species. It'll be interesting to see further developments...

Tue, 17 Dec 2002 15:04:36 GMT

Shop with Sherlock

I just released a little app for the Mac: Shop with Sherlock, a little shopping agent for the Sherlock application built in to Mac OS X 10.2 (Jaguar). It was fun and amazingly easy (especially compared with doing the same thing at Jango five years ago!) It's off to a good start - I have several thousand users already.

Thu, 21 Nov 2002 18:42:58 GMT

Mac vs. Windows

I've been following the debate between Steven Den Beste and Brian Tiemann with interest. Steven makes the quantitative argument that due to the high capital cost / low marginal cost economics of both software and chip design, Apple is just not going to be able to keep up. Brian counters with a variety of emotional, anecdotal, and qualitative arguments. I think they're both right in their own frame of reference, but I think the key factor is something else.

Most people who use the phrase "disruptive technology" without reading Christensen's book seem to think it means a huge leap forward. In fact, his original meaning is almost the opposite: As a technology market matures, the industry leaders follow their most demanding customers past the point where the market continues to value marginal improvements in technology. This opens the opportunity for disruptive new entrants at the "low end" with products that are inferior by all the standard performance metrics but deliver superior value along some other dimension.

I think the desktop market reached this point a while ago. First of all, software: I've been developing in Windows since version 2.0, but several years ago the emphasis clearly shifted and my primary technology stack has been internet protocols and infrastructure, with the desktop as a minor component. About a year ago my Compaq laptop had accumulated enough DLL/registry cruft that it was due for its semi-annual Windows reinstall, and I decided just for the heck of it to install Linux instead. I had StarOffice, plus Windows running in VMWare in case I needed it. It was an interesting experience, and it became apparent that the latest Linux distributions provide enough usability and functionality to compare very well with the Windows 95 era of desktops - which many users still run quite happily.

After a few months I got tired of fiddling under the hood of the OS and fighting atrocious user interfaces. I had come to like having my familiar Unix tools to work with, and I wanted to edit DV footage of my baby, so I bought a used PowerBook G4 on eBay. I'm still quite happy with it. It does everything I need, with an aesthetically elegant user interface and industrial design. I fire up Virtual PC a couple of times a month to run a version of Quickbooks compatible with my accountant's, and I use my old Compaq to run Visual Studio .NET, but otherwise I'm just fine. I spend most of my time in the web browser, email client, Office, bash, Palm Desktop, Eclipse, and Emacs. There are a few applications I miss - like Ecco! - but some that I would really miss if I moved back - like OmniGraffle, which is even better than Visio for diagramming. Overall, there's very little disadvantage to me from being on a non-mainstream platform.

I'm not anti-Microsoft - in fact, I think they're one of the best-managed corporations in the world. But they're facing a challenge. They have a lot of brilliant people and they are sure to survive and even thrive, but the reality is that they're no longer the center of the computing universe like they were 5 years ago. In a competitive market, the equilibrium price of a good tends to its marginal production cost, which for software is zero. There are strong signs that many parts of the software industry are headed in that direction.

On the hardware side the story is similar. It's quite true that Macs are underpowered relative to PCs, but it really doesn't make a difference to most people once they put in enough RAM that it doesn't swap. I'm running a 550MHz G4, and I was sorely tempted by the new 1GHz TiBooks. But I've already upgraded to 1Gb of RAM and a 60Gb disk, and doubling my processor speed would make essentially no difference to my daily experience. Likewise, my old 500MHz P3 Compaq laptop was just fine with 640Mb RAM. When I really need computing power, a $100k Linux cluster of Athlons gives me Cray Y-MP class number crunching.

So I think if Apple can get their prices down to a level where the overall package cost is in the ballpark, which they seem to be doing, then they can be a very reasonable choice for many people. It's now more like choosing a BMW over a Toyota, not like choosing an amphicar over a normal vehicle.

Afterthoughts:

It's instructive to read John Walker's prescient 1993 essay on subscription software, Programs are Programs, along with his 1997 prediction Microsoft at Apogee.

Microsoft's attempts to stave off the spread of open-source software in India. This is serious. If India & China decide that Linux is good enough and cuts 30-50% off the cost of (legally!) setting up a productive desktop, then the network effect suddenly flips the other way and 85% profit margins on Windows are ancient history.

Wed, 13 Nov 2002 13:52:55 GMT

The Eighth Day of Creation

This book by Horace Judson is one of the best science books I have ever read. I've been recommending it to almost everyone I've met since I started reading it. It is a gripping and enlightening tale of the key discoveries that are the foundation of modern molecular biology - the role of DNA in genetics, the DNA transcription process, and the role of structure and function in proteins. He does a masterful job of putting you in the time and place as it happened, with the key scientists and their ideas, inspirations, and false steps.

Wed, 13 Nov 2002 13:42:27 GMT

Understanding Windows EAL4 Evaluation

An amusing quote from Jonathon Shapiro's analysis:

"Security experts have been saying for years that the the security of the Windows family of products is hopelessly inadequate. Now there is a rigorous government certification confirming this."

Unfortunately, that's more security assurance than we have about any other commonly used operating system.

Tue, 15 Oct 2002 16:25:02 GMT

Experimental economics

Great interview in Reason magazine with Vernon Smith, recent co-recipient of the Nobel Prize in economics. Another great person at GMU is Robin Hanson. I had dinner with him when I was at IEEE Nano, the first time I'd seen him in several years. We had a great conversation about the applications of idea futures to corporate decision making, and scoring mechanisms for complex idea futures markets.

Mon, 14 Oct 2002 17:54:56 GMT

Monday morning humor

Via David Isenberg's SMART letter, some useful software features.

Wed, 02 Oct 2002 19:23:14 GMT

Lou Gerstner's farewell letter

Via John Ludwig, the full text of Lou Gerstner's farewell letter to IBM. He talks about the "massive changes underway in the industry", and how he reorganized IBM around customer focus, open standards, collaboration, and technology leadership. Inspirational and insightful.

Tue, 01 Oct 2002 23:24:08 GMT

The way to groove

Ray Ozzie published his original notes on the market opportunity for a collaboration product that became Groove. It's a great document because of its focus on the individual and the reality of how people work and collaborate on a day-to-day basis.

Wed, 25 Sep 2002 17:34:36 GMT

Platform dynamics

Two good recent blogs on platforms - Joel Spolsky and Ray Ozzie. One of the hardest issues is how to balance developing the platform itself and enabling the ecosystem.

In Platform Leadership, Cusumano and Gower have some interesting history about the development of platforms strategies at Microsoft, Intel, Cisco and others. One of the interesting details is the degree to which internal organization is important. The appropriate balance between internally developed complements and supporting external developers needs to be clearly supported by the internal organization as well as the external actions of the company - there were a lot of conflicts between Intel Architecture Labs, for example, and various product groups in the company.

Thu, 12 Sep 2002 17:22:28 GMT

Hiding in plain sight

Who Benefits Most from the High-Tech Revolution?. David Wessel writes about one of the secrets of the new economy: the principal productivity gains and cost reductions are found not in IT-making but IT-using industries...

This is one of those things that's obvious once you see it - you wouldn't expect the producers to capture a significant fraction of the consumer surplus, especially in a highly competitive industry. But the end-users are much less visible, so the availability heuristic leads you to think they are less significant.

Thu, 12 Sep 2002 15:17:53 GMT

The King's Shilling

I wrote this essay quite a while ago about the idea of a constitutional amendment to address the inherent conflict of interest in government activity. I've mellowed since, and I'd write it differently now, but it's still an interesting notion...

Fri, 06 Sep 2002 00:55:25 GMT

Open Source Business Models

I went to a big dinner last night in Los Altos. I went in part because Greg Benford was going to be there - not so much to meet him (in fact I didn't talk with him at all), but mainly because his presence meant that a lot of other people I knew would probably show up. And they did - he was a good Schelling Point!

Mark Miller is trying to figure out how to make a business of Combex. This is the company he and Marc Stiegler have built around the E language and capability security concepts. The core of E is open source, in part to build a developer community to get the ideas out there, and in part to have the trusted computing base of the system transparent and verifiable, and that part has worked out pretty well. The company is subsisting on a trickle of DARPA money for now, and the question of the day is how to grow the company:

The IBM model: Sell consulting services and custom software development. Viable, but gross margins are low and fixed costs are high compared to selling software.
Sell a proprietary application on top of the technology - E would be great for something like Groove, with a truly secure application deployment model.
A mixed open/closed source model, either by licensing or patents. Possible, but tough to balance.

Offhand, I'd pick some combination of the first two, but then there's the really hard question: What application area to focus on? Where does it have the most comparative advantage?

Groove has a 2-year lead as a secure P2P collaboration app, but they've committed themselves very heavily to the Windows desktop, so there are probably significant opportunities left open - cross platform, specialized devices, etc.
Distributed games - this was the application that the original E was built for, at Electric Communities
With the proliferation of mobile devices - 802.11, Bluetooth, PDAs, phones, laptops, etc. - there may be an opportunity to come in as a standard infrastructure piece for synchronization or collaboration. But these environments might not be able to afford the speed & space cost of E as it stands now...
Running other people's code on your data - coordinated web services that can securely pass closures. Intriguing, but what would it be used for? Would people really trust it?

E has the potential to be so much better than the state of the art in security that it's hard not to keep thinking about how to apply it...