Subject: Re: Economics of software distribution
From: ghost@ALADDIN.COM (L. Peter Deutsch)
Date: Thu, 18 Mar 93 18:37:55 PDT

> Let me apologize in advance to Peter Deutsch, who is the author of 
> a very valuable piece of free software, for the following message.

No apology necessary; I didn't feel I was being attacked personally.  And
by the way, there are *two* Peter Deutsches who are the authors of
valuable pieces of free software: the Canadian, Peter Deutsch (archie),
and the Californian, L. Peter Deutsch (Ghostscript).

> 			[Flame on]
> 
> This is very unfair to the author of compress.  It was never intended to
> run on anything but Unix.

Yes, originally, but not in this case.  The source code I got had quite a
bit of conditionalization and miscellaneous hackery that was explicitly
and clearly designed to make it work on MS-DOS.  It just didn't work.

> Much free software is better written and better supported than much
> commercial software.  I find that:
> 
>    o commercial software doesn't work.
>    o I don't know why. (because I don't have source), and 
>    o I can't fix it.   (same reason)

I find that:

	- Commercial software mostly works, because if it didn't, the
company selling it wouldn't stay in business.

	- When it fails, I don't know why, and I can't fix it, but support
is part of what I'm willing to pay commercial prices for.  If I don't get
good support, or if I need it more than once or twice a year for a given
piece of software, I start looking around for alternatives.

On the other hand, I do share your frustration with commercial software
that *almost* works.  Even if I were going to be exclusively in the
commercial software business, I'd definitely offer source code to my
customers, either free or cheap, with some kind of reward for returning
bug fixes to me.

> Free software works no worse than commercial software, and I can fix it
> if I have to.
> 
> Maybe I'm just a crazy hacker; but I'd rather get
> something that I can fix when it's broken than something that I can't.

So would I, if it isn't too broken, if the source code is reasonably well
written, and if I can rebuild it with the development tools I have at
hand.  In the case of `compress', it was obviously written in the old
school of "throw in another #ifdef every time you run into a new
platform."  (I'm beginning to think I should teach seminars on how to
write effortlessly portable C code, because I see the opposite so often!)
Otherwise either I want the vendor to fix it, or I'll go elsewhere.

> My DOS development system crashes about twice a day.

Mine probably crashes oftener than that!

> Is Linux worse?

Well, now that you mention it, I haven't even been able to install Linux,
because its install program doesn't work with the SCSI disk I have.  I'm
only mildly annoyed about this, because I'm quite productive in the MS-DOS
environment (which is where I've done 99% of Ghostscript development), I
trust Adam Richter and collaborators to fix the problem, and after all, it
was free.

I'm sure Linux will be a lot better that MS-DOS when I get past this
problem, but that's comparing apples and oranges.  How does Linux stack up
against SunOS?  The only thing I've heard about it so far is that you may
have problems using it on a TCP/IP network, because its network software
is not really compatible with the Berkeley implementation.  If that's all
that's wrong with it, I'll feel I'll have gotten a great bargain -- at
least, until I start doing enough networking that I'll want to get a PPP
connection....

> The "problem" with free software is how to set up a system to encourage
> people to do it and how to get them paid for it.  Did Peter submit his 
> mods to compress to distribution?

What distribution?  Free software often floats around the net with
absolutely no indication of where to send fixes.  The most recent comment
in the copy I got was dated 1985, with no indication of which of the 6
listed authors is taking responsibility for distribution and fix
integration.

> Probably not.  Probably because:
> 
>    o It's a hassle.

Not if there were an e-mail address for the author.

>    o There's not much in it for him.

On the contrary, I'd be delighted if there were someone taking ongoing
responsibility for integrating bug fixes and improvements, which I could
then get the benefit of.  My experience doing that for Ghostscript has
taught me how much time it takes, so I appreciate it when other people
take on that task.

>    o Even if he wanted to do it, there is no way to incorporate the
>      mods into the main release of compress, where it would do the
>      most good.

See above about e-mail addresses and responsibility.

> What do free software vendors sell?  I'd guess these things, all of
> which are called support.
> 
>    o Hand holding

I don't WANT hand holding, I want software that works WITHOUT hand
holding!  That was a big part of my point about commercial software.

>    o Custom modifications

I rarely run into a situation where I wanted custom mods badly enough to
pay consulting rates ($50/hour and up) to get them developed, but it does
happen from time to time.

>    o Quick bug fixes

Why should I expect any better service from free software supporters
than from commercial software vendors, unless the former are using
the subsidized distribution mechanism (Internet) that the latter are
explicitly forbidden to use?

>    o Early access to updates

I don't want them until they work.

> The difficulty that vendors of free software have is getting customers
> to pay for what they *thought* they were getting for free, namely all
> the above and possibly more besides.  (Additions are requested.)

There's also an issue of visibility.  In the case of commercial software,
it's one-stop-shopping: if you know where to get the software, you know
where to go for support.  In the case of free software, you have to hunt
around for someone who's willing to support it, unless (as rarely seems to
be the case) the author offers support.

> As an example of how customer expectations vary, consider the difference
> in the way that chip vendors and software vendors handle, or at least
> used to handle, bugs in their products.  Chip vendors refuse, or at least
> refused, to admit that there is ever anything wrong.  On the other hand
> Sun publishes a magazine of known bugs and fixes, which the customers
> subscribe to at great expense.

I like the idea behind this observation, but I don't think it carries very
far.  Sun can get away with this because they are mostly living in the
Unix world, where quality standards for released software are lower, and
because they don't sell application software to end users, who demand
better.

L. Peter Deutsch :: Aladdin Enterprises :: P.O. box 60264, Palo Alto, CA 94306
ghost@aladdin.com, ...decwrl!aladdin!ghost ; voice 415-322-0103 ; fax 322-1734
	    "Implementation is the sincerest form of flattery."