Subject: Re: ok, FSBs created schwag; now bet the farm on something better
From: Peter Wayner <>
Date: Thu, 20 Sep 2001 08:58:28 -0400

 Thu, 20 Sep 2001 08:58:28 -0400
At 9:10 PM -0400 9/18/01, Jonathan S. Shapiro wrote:
>  > ...But they've also produced Microsoft Word, a
>>  software program with more features than anyone can ever use in their
>>  lifetime.
>Just a reminder: Word didn't come out of research. It came out of
>development. The research editor that it is (now distantly) based on came
>from Xerox, and was a thing of great beauty.

Okay, I'll bite. I was trying to brief, but I've got time here. The 
distinction is beyond argument because it driven by style, taste and 
class. One man's beautiful research is another's grim development.

It's very easy to find examples of how Microsoft "research" helps 
products. Here's a quote from an MS press release:

Examples range from ClearType® display technology and grammar 
checkers to functionality in Microsoft SQL Server that enables the 
high-speed analysis of complex data patterns. Recent products such as 
the Windows® 2000 operating system and Office XP include numerous 
underlying technological improvements developed by Microsoft 

If you read through the MS research website, it's easy to find 
"researchers" who seem to be glad to talk about how their "research" 
added features to Word and other MS products.

* If you click on a link to an audio file, you get the Windows Media 
Player with high grade crypto from the research group helping stop 
nasty copyright infringement. (That's a feature in demand among the 

* If you ask your .NET version of Word to use voice recognition, look 
for technology from the Speech Technologies Group:

* Do your True Type fonts look nice on the screen? Thank the graphics group:

There are dozens of examples that range from spell checkers to that 
pesky talking paperclip.

My point, though, is that much of the high-profile free software was 
not built by people seeking to make something "new and improved". 
Researchers in universities need to create "new and improved" things 
because the papers won't be accepted if they aren't new and improved 
enough. Marketeers need new and improved to get people to pay for an 
upgraded version of the software.

Large, important parts of the world of open source/free software is 
not really driven by "new and improved". Eric Raymond likes to say 
that it is built by people who want to "scratch an itch". That is, 
they just want the machine in front of them to do what they want. 
They're not searching for papers. They're not searching for jazzy 
features. They're not coming up with metaphors about agents, fuzzy 
logic, contraint-based reasoning or whatever the new academic fashion 
happens to be. They're just making the things run.

One of the things that continues to astound me about the 
Linux/GNU/FreeBSD/Apache world is how disconnected it is from the 
university. Long ago, Richard Stallman, Kirk McKusick, Keith Bostick, 
Jordan Hubbard,  and many others worked in universities. Not anymore. 
Linus Torvalds was never more than an undergraduate. Yet, the 
software is pretty nice.

The open source world is really a very different community that has 
similar goals to the academic world, but a style all of its own. 
Papers, citations, and to some extent grants aren't really worth as 
much. People write the software because they need it and they share 
it because they think that others might be able to help them.

Of course, now that I say that I should qualify it by saying that 
nothing in this statement is solid, guaranteed, or true for all 
members of the class. There are people in the open source community 
who say they do "research", there are people who say they do 
something entirely different. There are projects in the university 
that concentrate on producing usable code. There are open source 
projects that are infatuated with adding new features. There are 
plenty of counterexamples.

It's an argument about word choice and those discussions usually 
don't converge. I think, though, that the distinctions are important 
because the feedback loops are so different.