Software distribution the cypherpunks way? Was: Patent Fight Could Add to Cost of Inter

["Phillip M Hallam-Baker" <hallam@ai.mit.edu>]

This turned into a long article on software distribution.

> NEW YORK (AP) _ A little-known patent could raise the cost of 
> doing business over the Internet for companies selling software, 
> video or other digital products delivered online. 
> 
> E-Data Corp. of Secaucus, N.J., is suing 17 companies, including 
> McGraw-Hill and CompuServe, to collect licensing fees on the 
> patent, which protects downloading of encrypted digital 
> information. A court hearing is scheduled Dec. 6 in New York on the 
> company's claims. 

Although this is one of the most bogus and indefensible patent claims
about I don't think it has any material bearing on the Web. Any that
it does have is likely to be beneficial since the Freeny model is 
inherently broken and non-Web.

What Freeny does is to claim the idea of encrypting an information 
product, delivering it through a computer network (though not a
physical medium, there is another patent that claims that), and selling
the decryption key rather than the product. It is a hot idea if you are
trapped in traditional, pre-Web thinking but not if you understand
the dynamics of the net.

The problem is that encryption is not such a great barrier to 
preventing proliferation of software. If encryption is free than you 
might as well add it in, if there are patent costs then you have
to ask what precisely you gain from encryption. The problem is that 
the information good has to be decrypted before use. Since it is
very difficult to prevent copying after the decryption has occurred
all that the encryption is really doing is to slow down the process
of proliferation.

A more important question is what if anything one gains from 
preventing proliferation. Microsoft has not been hurt by the massive
bootleg copying of its software, far from it. Many business users
are keen to buy non copy protected software because they can then
obtain a "free" copy for their home machine at no cost to 
themselves. This minor peculation is in my view the principle reason 
for the dominance of the PC in the home market. There is no good
reason why a person would buy a $2000 business machine rather
than a $500 home machine when the latter plays much better games.
But add the idea that one can use the software bought by one's 
employer and the PC looks the better buy.

I believe that the success of Netscape demonstrates that copy 
protection is a bad idea. If you believe in free markets then 
the price of any given product is going to come down to the marginal
cost of production. For software sold via the net the marginal
cost is pretty close to zero. The price of the various "Office" 
bundles sold via the stores is astonishingly low. $200 for MSOffice
or Corel Office is a silly price. If the product is bundled with
the machine the price gets even sillier - a few dollars. I expect
that before long we will see machines coming bundled with more than
one office suite because the prices are just too low to make the 
choice irrelevant. 

An important fact to bear in mind is that companies are in general
interested in being honest. There is little incentive for an 
employee to cheat the software vendor. The cost of lawsuits is a
good deterrence since the type of company where management are
likely to encourage software theft is also the type of company
that is likely to have disgruntled ex-employees willing to inform.
As a system manager my principle interest in net O/S such as 
WNT and VMS is that they have good, centralized and standardized
licensing mechanisms that allow me to make sure that the software
has been paid for. Its another reason why site licenses are useful!

The price of net software is likely to be close to the price
the OEM pays, a few dollars. This is likely to be structured
somewhat differently however. I expect to see growth of two 
trends, software rental and componentware.

Componentware, selling the package in small increments is a
poor proposition if selling via physical media. The cost of
a CDROM is low, (few cents), but the cost of post and packing
is high. Selling componentware via the net is considerably more 
interesting. Just as I paid a few bucks for my XV program,
I'm willing to pay a couple of bucks for additional input 
filters. I'd certainly be willing to pay a couple of bucks
for specialized document tools or specialized canned document 
formats (e.g. IDEF0, ESA-Requirements whatever). I see the
potential for componentware being primarily low cost, high
volume. 

Componentware is now on the brink of being practical for three
reasons. First as previously stated the net makes distribution
costs lower. Second modern software applications are considerably
more "open" to extension than previously. In the past most
software houses attempted to produce products that satisfied every
need. Today the large software houses realize that they cannot
hope to satisfy every need but that if they can cover 95% of
a persons needs in a way that allows them to add in the remaining
5% everyone is happy. Finally, Java has legitimized the 
componentware idea. Its worth noting that the technical suitability
of Java to the task is irrelevant. What is important is the fact
that several hundred startup companies have been created to do 
nothing else than make Java applets. Such companies are effectively
committed to make componentware work and it is inevitable that
at least one will succeed in establishing a paradigm for the
others to copy.

The second interesting development is software rental. To 
an extent this already occurs since the rapid rate of software
development means that software rapidly becomes obsolete. Anyone
who is using WordPerfect 3.1 for more than a few hours a year
should probably upgrade to the latest version simply because it 
offers so much more functionality. The recent move by Microsoft
to incorporate the year into their products is intended to 
encourage frequent software upgrades. It is a small step from
this model to one in which software is "leased" for a period of
a year or more at a time.


Of course both componentware and rental strategies have been 
around before the net existed. What the net does is to change
the economics so that low cost, high volume is the mainstream
strategy. Software rental traditionally involves expensive
agreements, typically brokered by lawyers on both sides. The
network means that it will become economic to rent software
in low volume and for short periods.

Joining the componentware and rental strategies together would
point to a distribution strategy in which a corporation would 
lease a certain number of seats worth of fully updated packages.
The seats would be continuously updated on a daily basis as 
additional components passed the supplier's QA.

In this model the purchaser is buying the up-todateness 
of the package. There would be little value in bootlegging
the software since its unlikely that the bootleg version 
would be anything like as convenient.

There would also be scope for different levels of uptodateness.
Just as the Microsoft developer network delivers a bulging binder 
full of CDs to my desk each quarter there will inevitably be a 
large number of developers willing to risk buggy software to 
have access to the cutting edge technology. On the other hand 
there will be people who prefer to wait until others have found 
the bugs, waiting a judicious length of time before updating
unless a patch is a response to a known bug.

Encrypting the distribution process might be worthwhile to 
reduce fraud but the main vulnerability is probably
unauthorized copying after distribution. 


What the Freeny patent covers is thus the old model of software
distribution in the new sales paradigm. There is little point in
taking costly measures to protect the copyright of products 
being sold for small dollar amounts. First Virtual proves this.

Where more expensive software is concerned it would ideally be
best to use the net. But here the price of the patent is simply
too much. The 1% extorted under the patent would easily pay
an intern to stuff a CD in an envelope and send it.


The role for encryption in the software distribution process
is in authenticating the delivered goods. The role of encryption 
is to provide privacy. It is very hard to keep a secret that
every one knows about, still harder to sell information  that is
public knowledge. The net is not an information economy, it
is an access to information technology. Nobody will profit
from merely selling information, they will have to sell 
the ability to access information in better ways, better indexing,
faster access, lower publication delays and so on.

comments?

		Phill

PS in checking this document Microsoft's spellchecker suggested
"Mosaic" when checking the word MSOffice!