Subject: Re: Support as insurance
From: Ian Lance Taylor <>
Date: 29 Nov 1999 12:30:23 -0500

   Date: Mon, 29 Nov 1999 10:19:34 +0000
   From: Ben Laurie <>

   Bob Young wrote:
   > Whether or not your product actually is more reliable or higher value
   > than the alternatives available in the market is besides the
   > point.  Customers will buy -more- warranties from you for the higher
   > quality software you sell - if, and only if, you can convince them that
   > you are a more reliable supplier of solutions to their business
   > problems than your competitors.

   This reminds me of a rather depressing study done by (I think) IDC in
   the early 80s that aimed to find out what was correlated to software
   market penetration. The only factor with any significance was marketing

I'm not sure why this is depressing, or, for that matter, surprising.

The experience of Linux, gcc, Apache, sendmail, bind, etc., shows that
a superior solution can achieve good market penetration with zero
marketing budget.

When the available solutions are more or less equivalent--as appears
to be the case among, say, spreadsheets or word processing programs or
Linux distributions--then marketing is the main determining factor.

In a fantasy world of perfect markets and perfect information, this
would not be the case.  But in the real world, choosing among
competing alternatives means investing just enough time to learn which
alternative will do the job.  Since for most people software is a tool
to achieve other ends, rather than an end in itself, most people just
confirm that one of the software packages they know about will work
for them.  Which software packages they know about is a function of
marketing budget.

Another way to put this today is: don't go head to head against
Microsoft.  Or, unlike most cliches, ``build a better mousetrap and
the world will beat a path to your door'' is simply false.

The sucessful zero-marketing packages listed above share a couple of
other characteristics: they all have gratis distribution, and they are
all selected by people who tend to know a great deal about software or
system administration (compare to selecting a word processor).  For a
free software package to be widely chosen as a tool by people who do
not care about software, good marketing is required.  This is what Red
Hat is presumably trying to do to push a desktop version of Linux.

I do optimistically believe that superior products tend to win out
over time.  But the victory can take several years or even decades
(network effects in particular can ensure a long lifetime for
inferiority).  And the victory can take the form of the dominant
products adopting the superior features, with no monetary reward for
the people who introduced those features.