[
Lists Home |
Date Index |
Thread Index
]
At 08:29 05/05/2003 -0400, Mike Champion wrote:
>I had a very similar reaction to Michael Kay's response. But there's a
>third option, "make everything small enough for one person to do
>completely." That's more or less the idea behind modules which begat
>objects which begat components which begat services -- "small pieces,
>loosely joined."
Which simply shifts the emph to the interface.
Designing missile subsystems isn't too bad. a good processor, a few hundred
k of memory
and a headfull of code.
The pain comes when we stick n sub-systems together.
Hence the need for the interface design spec, definition, test spec blah
blah blah.
Michaels boundaries are him.
As soon as boundaries require shared understanding it all starts to fall
apart.
<grin/> Back to semantics. The interface farce is quite humorous. 'I
thought this meant
X.... No, it means Y. Ah sh..oot. That means I have to re-design my blurt
box to
provide you with data at rate X. I can't do that for the trials!
the interfaces are just as/more important than the 'headfull' of code.
regards DaveP
|