Wasting half a trillion dollars?

["W. E. Perry" <wperry@fiduciary.com>]

There is a full-page PeopleSoft ad in today's (Monday: The Information
Industries) business section of the New York Times. Above the fold, the
page is filled with the headline:

Last year,
code on the client
cost businesses half
a trillion dollars.

That is followed for the next quarter page with "How much is it costing
your business?" and then the following paragraph nestled at the bottom:

PeopleSoft 8. The only enterprise solution with no code on the client.
Every year, your company spends hundreds of thousands of dollars
maintaining software on each individual PC. But PeopleSoft runs entirely
on the internet, and that makes it accessible from any web browser,
anytime. So you can work collaboratively with your customers, suppliers,
and employees, in realtime, from anywhere in the world. Which opens the
door to a whole new way of doing business. Suddenly, "no code on the
client" is more than a strategy to cut implementation costs. It's the
key to running your business more efficiently and making your people
incredibly productive. And most importantly, it's exactly what you need
to create a profitable, competitive, collaborative enterprise.

At the bottom is the PeopleSoft tagline: "People power the internet."

This reads (at least to me) as if it were consciously designed as an
anti-advertisement; that is, I can hardly imagine a more perfectly
backward message. I will gladly take as a first premise--indeed, an
article of faith--that 'people power the internet'. Indeed they do, and
I am further pleased to see the small-'i' spelling of 'internet'. This
is precisely what internetworking is about:  local networks built to
local standards which perfectly express, and facilitate, the functions
of the individuals whose expertise is the raison d'etre of those
networks. To achieve 'best of breed' functionality we do not impose
standard processes, or standard data structures, on those who perform
those uniquely valuable functions. Instead, we interconnect with a
single addressing scheme these individually customised networks. This is
the internetwork (or Internet) topology, which is in place and
functioning well. Yet when the question is no longer 'how do we route
packets among these local networks (some as small as a single
IP-addressable node)', but instead 'how do we distribute processing (or
execute transactions) between these nodes', how can the answer suddenly
be that we must abandon the local autonomy that the internetwork was
designed and built to protect? Surely the 'internet' answer (and the
answer which best respects the primacy of people, as individuals) is
that we should route procedure calls, or data structures populated to
trigger processing, blindly, just as we do any packets, relying on the
recipient to do the right thing given its unique context and own
expertise. The perfect expression of that expertise and acknowledgment
of that specific context is, in fact, 'code on the client'. That code
may exist in only one place and may be designed and implemented on
principles and to standards seen nowhere else. It is the function of an
internetwork to make just such processing available to nodes which
because of local constraints of their own cannot be built on those
particular premises and therefore not implement that precise
functionality. This 'code on the client' is worth building because it is
the exact expression of the best particular expertise for a given
process. It is worth the expense of maintaining because it is a uniquely
valuable common resource--common because it is (at least potentially)
available to all through the internetwork addressing scheme, not because
it depends on a priori agreement on the standards to which it should be
built.

Respectfully,

Walter Perry