[
Lists Home |
Date Index |
Thread Index
]
That's sooo true. Every time we get a requirement
for an application, a local tells us we need to
build tools for the customer so they can build
the application they require and in some sense,
where customization for the locale dominates,
that is a good idea. But if you have to maintain
a core for all of these guys, it is a lot of work
and that is why you have to sell a lot of copies
to make a buck. It's like selling sheet music. The
printing costs alone can make the margins thin.
If you saw how many configuration options goes
into a single records management system for
police agencies, you'd puke. But... that is
what sells the dammed thing in a world where
we don't market shrinkwrap, but saranwrapped
apps. We are RFP-driven; not marketing driven.
Yet, XML application languages need a way to
convey semantics or interoperation predictability
goes to hell. We've seen this same problem
over and over again from the very first GML
apps forward. We worked on it in the MID. We
got killed by Yet Another Gencoded solution.
As soon as one gets beyond gencoding, this bugger
rears its ugly head. Someone eventually hardwires
to the codes and gets market share. We can end up
cooperating ourselves to bankruptcy. The guy who
makes the money is the one who watches all the
candidates, figures out which one has "just enough
mindshare" and "ride Sally ride!". It is true
that the early bird gets the worm but the second
mouse gets the cheese.
len
-----Original Message-----
From: Joshua Allen [mailto:joshuaa@microsoft.com]
Yeah, someone could say, "I am going to grow corn so I can ride the
markets for cereal *and* cow feed". Or, "I am going to go into corn
distribution so I have leverage in both directions". Or, "I am going to
go into retailing so I can get 50% margins on a box of corn flakes".
They are all correct.
A side opinion; I think computer people are the worst in the world for
always going meta. Even when we have a clear problem to solve (like,
implement an accounting system), we always try to attack the meta
problem and the meta-meta problems first. I admit that abstraction is a
good way to attack problems, but taken to far it is a way to avoid
solving problems while still looking busy.
|
