[
Lists Home |
Date Index |
Thread Index
]
And while we know most of this by rough experience, the
following is fascinating reading relative to Ken's post.
http://www.tdan.com/i030hy01.htm
What does and doesn't work. Worthy of some XML-Dev
permathread amplification of worse-is-better, but really,
discovery based systems work better than deep C2. The
airplane starts out as a kite with an engine. Then
spiral development/evolution takes over. As long as
one can afford the knowledge maintenance, this keeps
working. At some point, gaps start to show up in the
system because externalities (an economics term) begin
to cause costs to rise (see the airline/union lawsuits)
and then the system which is complex, therefore sensitive,
begins to fail. Eventually, a Saturn V would have blown
up just as the Shuttle failed and other non-manned systems
failed under the 'cheaper, faster, better' approach.
Network effects are peculiar that way.
CoT chopped it down to just the essential bits. That
works for awhile, or at least, it gets enough networked
nodes talking and doing useful work while the non-real time
powers negotiate the more complex transactions. It isn't
that You Ain't Gonna Need It but You Ain't Gonna Need It First
so Do What You Know Until You Are Sure of What You Don't Know.
Then it is, what can you afford? A very real reason for moving
to markup even before XML was attempting to solve the complexity
problems of maintaining large aggregate systems. If the cost of
the document system goes up with the cost of the system that it
documents, you have a classic strongly linked cluster problem.
For want of a nail, etc.
To answer Rick Marshall: simplifying XML will be better than
making it more complex. But the costs haven't been justified.
Just as some want to make it simpler, others such as Jakob Nielsen
are discovering that linktypes could be better
using all the tricks we knew 20 years ago before he started
filing patents.
The only question a poor man asks is 'who pays'? If the
network effect takes over, everyone does, eventually although
the first mover pays the most. But network effects usually
don't happen for complex designs and that is the winnowing
or Darwinian effect also known as churn. It's a piece of
cake to keep a 66 Beetle running if you can get the parts.
It's not so easy to keep a IBM 360 series running even if
you can get the parts. That is, value is not only money;
it is utility.
len
From: Bill de hOra [mailto:bill.dehora@propylon.com]
I'm with Karl and Len. I received a mechanical chronometer last year
from my family. It's hands down the finest object I have ever had, and I
fully expect it will outlast me. It serves a constant reminder about
simplicity and robustness. Paul Graham said once that you can buy a more
accurate and functionful modern watch for the fraction of the cost of a
something like a mechanical chronometer. I think that's true, but it's
also missing the point in some thing like the way meat market IT and
tools-will-save-us arguments miss the point.
|
