Re: Sharing Models, was: Intestinal Fortitude

From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Tue Mar 20 2001 - 00:55:35 MST


On Mon, 19 Mar 2001, Spike Jones wrote:

> Robert, we desperately need a leader with vision.

Here is how I would "direct" (at the risk of annoying everyone)...

Excel spreadsheets are to me "open source". Spike and I
share them without any problems. That they run on "closed
source" Excel doesn't matter. Excel runs on closed source
hardware which runs on a closed source universe.

Should you try to increase the number of platforms on which
things run? Yes. [Go look at my CV and see the number of
platforms I've worked on.]

So Excel spreadsheets can run with the open-source Excel
equivalents. The only problem is I do not know how
"equivalent" these are. What features of Excel cannot
be used in the alternate open source emulators? [If its
as small fraction then we try to keep the spreadsheets
from using those features, but we will need to document them.]

Now, Mike is web-interfacing Spike's sheets. That is great.
But will his approach work on the open-source excel equivalents?
If it will not, then I would suggest that it may be useful for us
all to up-educate ourselves to the more open source, portable
tools (Python, etc.).

While I appreciate and agree with Eugene's perspective, I have
to be pragmatic that the people on the list have limited time
to dedicate to sharing their work. Many things shared running in
a not insignificant number of locations is worth more than very
few things shared running everywhere.

I don't learn Russian to talk to Russian scientists -- Following
that principle I would have to learn a dozen languages to talk
to dozens of scientists. I say "if you want to work with me, you
have to learn some method of communicating with me". Following
this principle, you don't require the creators of the apps
to learn whole new languages to implement the apps. Let them
implement the apps (with documentation!!!) in what they know.
Its then its up to others to adapt those apps to more open
environments if the need is there and the multi-environmental
individuals who can do this are interested.

Remember the standard programming axioms -- the first version
is always a prototype. If its useful enough to make derivative
versions, *then* you think about how to reach the greatest audience
with those versions.

Really, how many people on the list want to do rocket optimization
experiments? 5? Do any of them not have access to Excel? Unlikely.

The bottom line is this -- Open source is great from some perspectives.
It isn't great from others. For example I have the source to my
email program "Pine". Can I get it to correctly sort my inbox?
No. Why? Because when you encode the sorting rules of the type
I require it corrupts the messages. Do I have the knowledge to
debug and fix this? Yes. Do most people? No. Have I fixed it?
No. Why? Because I've looked at it, added debug statements and
come to the conclusion that whatever it is doing wrong might require
several days to fix. It simply isn't worth my time.

Is a situation in which companies that have millions of paying customers
who encounter closed-source bugs and report them better or worse
than a situation in there are open-source programs that were developed
ages ago as a "public-service" but are now are essentially unsupported?

Eric Raymond is correct -- in the shadow of many eyes all bugs
are shallow. The very important question is *how many eyes*
(with sufficient wisdom behind them) are looking at the bugs?

While I have no love for Microsoft (Windows '95 & '98 are jokes)
and NT I can regularly bring to its knees (open up 60-80 Netscape
or IE windows), Linux is no better -- no journaling file system
(until recently), no "raw" devices (like the operating system's
"read-ahead" strategy is optimal for databases). BSD is better
because it is "real" UNIX and open source, but the *best* UNIX
is that that went from AT&T to Novell to SCO, which was closed
source, was a much better engineered system (at least while Bell Labs
still drove the cart). Novell Netware too was a much better server
in its heyday, despite its closed source, because the original
developers *knew* how to engineer an efficient system.

There are examples of open source software that is very very good,
such as Perl. I'd argue however that probably 60-80% of the open source
software out there is caca. I'll base that on the simple perspective
that there are not that many *really* good software engineers.

Robert



This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:59:41 MDT