Subject: Re: fixing bugs....
From: Keith Bostic <bostic@abyssinian.sleepycat.com>
Date: Thu, 30 Mar 2000 11:30:11 -0500 (EST)

> http://www.unitedmedia.com/comics/dilbert/archive/cal-35.html
>
> On the one hand, this could be about Keith's concern that fsb's have
> an incentive to insert bugs.  On the other hand, it presumes that the
> bug-fixer has a monopoly.  That they don't is precisely one of the
> reasons why fsb's have an advantage.

Where the product is reasonably complex, or the customer reasonably naive,
the bug-fixer has a practical monopoly, and the "fact" that the customer
can fix their own bugs is largely an illusion.

Customers are rarely competent to fix their own bugs, and would rather
pay someone to do it for them.  As it's less expensive to pay the company
that wrote the software to fix it than to fix it yourself or hire outside
help (because of start-up costs and the costs of re-integrating the fix
on each new release), bugs can make money for a software company.

I'm a competent software engineer, but when I'm acting as a customer, the
last thing I want to spend time doing is fixing someone else's software.

We use lots of free software in our development cycle: gcc, autoconf,
libtool, weblint, Tcl and more.  We will do almost anything to avoid
having to work on those tools.  I will eagerly pay someone else to do
any work we need because it's a waste of my engineering resources and
detracts from the company's focus.

As an analogy, I generally pay 20-30% more than I have to when I buy
hardware.  Why?  Because I know when it arrives it will boot and run,
and I won't have to waste time on because the VAR shaved $3 by putting
in a cheap Ethernet card that sometimes crashes when the load is high.

To be fair, this whole argument is based on money being easier to find
than engineering time, but I believe that's the case for most of the
industry.

It's also based on the belief that few companies will throw out an
approach that "almost" works, because they think there's just a few
more bugs to clean up, and so they end up with a recurring support
cost instead of ignoring their sunk cost and re-evaluating the problem.

Finally, well, OK, but remember, you brought the subject up, I didn't.
One of our sales guys sent around the following email last month:

    From: XXX YYY <ZZZ@harvard.edu>
    Subject: Re: Berkeley DB support

    I am sorry to say that we have decided not to purchase a support contract
    for your Berkeley DB software.  It was thought to be too expensive for a
    product that works so well.

--keith