[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Re: Can a machine test itself? + other issues

Dick Johnson says:

> The economic basis of computers was and is that they may run
> more than one program.  It does not have to be this way; you
> could design and build a dedicated, one-purpose machine.  Quite
> expensive, really, since all development costs must be amortized
> over one specialized application.  The profit potential is limited,
> given the costs of development, so no private party would want
> to do this.

The evolution has actually gone the other way -- given the low
cost of microprocessors, it's not necessary to avoid the "waste"
of dedicating them to a particular purpose.  So to lower development
cost, lots of dedicated, one-purpose machines have characteristics
that formerly defined computers (general-purpose CPU, stored program
written in a high-level language, etc.).  The characteristic they
usually *don't* have is running more than one program.

As a concrete example, I'm making a great living doing software
for wireless networking equipment.  Our practice is quite the
common one in the industry (from startups to giants like Cisco
Systems), so I'm not giving away any secrets with this description:
It's programmed mainly in C, and has some capabilities you'd
recognize from Unix or NT, but has no hard disk, and no easy way
for the customer to write, load or run any program except the one
we embed on it.  The hardware is a mixture of general-purpose CPU
and special-purpose (but nonetheless mass-produced) network switching
chips.  As to the OS, there's a little real-time executive in it,
bigger than an old micro system like CP/M, but certainly nowhere
near the size of a something like NT or Solaris.