Colloquium at Stanford
The Unfinished Revolution

Memorandum


Date: Thu, 2 Mar 2000 01:33:35 -0800 (PST)

From:   Eugene Kim
Reply-To: unrev-II@onelist.com

To:     unrev-II@onelist.com

Subject:   How DKR Penetration Will Be Achieved

Interesting post, Eric. Couple of comments below.

On Sat, 26 Feb 2000, Eric Armstrong wrote:

There are several reasons that prevent this approach from being viable. Chief among them are:

  1. Organizations simply do not work that way. While some seriously desire to improve their productive capacities, virtually none want to "improve their capability to improve".

I'm not sure I agree. When a company tries to facilitate communication among its employees by building centralized cafeterias, intranet communities, or even rearranging cubicle arrangements, are they improving their productive capacities, or are they improving their capability to improve? I would argue the latter. Building a centralized cafeteria, in and of itself, is not going to improve a company's productive capacities.

Another thing to keep in mind is Jeff Rulifson's point, which was that having A, B, and C activity in an organization does not in and of itself make that organization a bootstrapping organization. I think this distinction is central to Jeff's bootstrapping the bootstrapping proposal, where he drew distinctions between Type I and Type II organizations. In some ways, Eric's post reminded me of Jeff's proposal.

[Quoting again from Eric's letter...]

  1. To understand why such a project is anathema to management, it must be understood that the risks are huge. First, the cost of failure is high. And, to succeed, it will likely require change to the organizational model. Such change is always risky, and usually resisted by lower echelons who perceive it as "interference". That makes the rewards highly uncertain. And even if the rewards accrue, the payback period is so long, and the results so far removed from the source, that there is a serious danger that the contributions will not even be recognized.

Curtis Carlson made a persuasive argument at one of the sessions for why this thinking is not necessarily the case. It's a question of dimensionality. If you go to management and promise a 20 percent increase in productivity if your very expensive, very risky proposal succeeds, then you're right; management will most likely laugh you out the door. However, in today's economy, something that promises a one percent increase in productivity with a high probability of success will compound very quickly because of shortened product cycles caused by Internet time and the greater importance of capturing market share quickly.

Here's an example from the world of free software. A few years into the Linux project, Linus Torvalds (the creator of the Linux kernel) decided that he would port the kernel to C++. That never happened for a number of reasons, one of which was compile-time. At the time (around 1993), C++ compilers were considerably slower than C compilers. Now a 10 minute differential doesn't seem like a big deal for software that takes an hour to compile, but for kernel developers who are constantly recompiling, it makes a huge amount of difference. This problem, in fact, is one of the driving factors behind research on incremental compilers, much of which is happening at IBM. The difference between recompiling a file and all of its dependencies versus incremental compilation may be a matter of seconds for a single instance, but compounded, it makes a big difference in productivity.

How does this apply to DKR penetration? I think this was the very basis of Adam Cheyer's proposal. Start small with what's available, achieve small but measurable improvements, and build on that.

Notice that there's a large amount of overlap between my points above and most of Eric's points. What I'm suggesting is that maybe Eric's proposal isn't as orthogonal to Doug's ideas as one might think.

[Quoting again from Eric's letter...]

Where a paradigm shift like the computer is concerned, penetration into executive ranks has been remarkably slow, presumably due to the amount of training required.]

This is somewhat irrelevant to the broader issues, but just for the record: Perhaps executives have been slow to adopt computers simply because computers do not make them more productive.

Sincerely,

-Eugene

Eugene Kim
eekim@eekim.com
http://www.eekim.com/