Subject: Re: street performer protocol
From: Ian Lance Taylor <>
Date: 23 May 2000 09:04:00 -0700

   Date: Tue, 23 May 2000 06:55:58 +0000
   From: Crispin Cowan <>

   > I don't deny that competition will reduce unit costs and make it
   > easier to switch to other firms' versions of the same product.  But
   > the switching costs of changing technology do not decrease.

   The FUNDAMENTAL point to using open source products is to avoid having to
   switch technologies when you switch vendors.  If you repeatedly insist on
   blowing this effect off, then you're going to continue to miss the point.

I see two main cases.

One is where there are multiple FSBs supporting the same product.
Switching vendors without switching technology is easy.  There can be
significant cost competition among vendors.  Running the FSB
profitably can be difficult, as margins can be low.

The other is where there is one FSB supporting a product.  Switching
vendors without switching technology is difficult.  Granted, having
the source makes it possible to start supporting the code yourself.
It also makes it possible for somebody else to start supporting the
code.  But both alternatives would be costly.

Buying a libre product from a vendor which might go out of business is
not obviously better than buying a proprietary product from a vendor
which probably will not go out of business.

Also, the fact that there is one FSB supporting a profitable product
does not mean that another FSB will come into existence supporting
that same product.  The second FSB will have to invest in learning the
technology and will have to compete against the significant market
advantage of the first FSB.  The first FSB can keep their profit
margin low enough to make this unattractive for any potential second
FSB, while still making enough money to continue in existence.

   > However, the competition that results from multi-sourcing necessarily
   > reduces the surplus available for future development of that product
   > (compared to a monopoly),

   But if the class of products is open source, then the workload is shared
   across the vendors, and the result is a wash.

It can happen that way, but I don't see it as a particularly common
occurrence.  I think a more likely arrangement for a product with
multiple FSB supporters is a couple of leading companies which make
significant investment and a number of coattail riders who make
little.  What is the comparable investment of Red Hat and Mandrake,
for example?  (I don't know the answer).

This effect is why Red Hat is not going to succeed on the sale of the
vanilla boxed product.  I expect that those boxes make them very
little money overall, and they mainly do it to promote brand awareness
and to mop up the odd dollar.  Fortunately Red Hat has other sources
of income.

   >  The point is that
   > switching among suppliers of essentially identical products is _not_
   > the kind of switch that software-related managers generally worry
   > about---

   No; managers worry about forced switches that occur when their vendor goes
   bankrupt.  Open source makes this a non-issue.

No, it really doesn't.  As I said above, if I buy a libre product from
a vendor who goes out of business, I'm almost as screwed as I am if I
buy a proprietary product from a vendor who goes out of business.
It's true that I have options in one case which I don't have in the
other.  But those options are extremely unattractive.

   Rather, the big issue with proprietary software is "how soon will the vendor
   make me switch to a churned version that I don't want/need to switch to?"
   E.g.  MS is going to discontinue support for NT4 in 2001, more or less
   forcing customers to adopt W2K, much as they forced everyone to switch from
   Win3.1 to Win95.

   In contrast, open source vendors can't force customers to do much of

Sure they can.  Red Hat can force people to upgrade too.  I think they
actually have dropped support for releases before 4.2.  The only
difference between this and your Microsoft example is that it is
possible that an aftermarket will arise in supporting old Red Hat
releases, whereas it is impossible for there to be an aftermarket in
supporting old Microsoft releases.  But if that aftermarket does not
actually arise--and I see no signs of it--then customers who are not
in the business of supporting old software themselves are forced to

It's true that Microsoft can do things that make people scream, and
Red Hat generally can not.  But even Microsoft can't make people
scream too often, and they would be even more restricted if they
weren't a monopoly.