Re: APIs, messaging

["Simon St.Laurent" <simonstl@simonstl.com>]

On 23 May 2001 02:26:34 -0400, W. E. Perry wrote:
> Of course, I want to push the point even further and insist that the 
> software of close tolerances is viable only on the homogenous network 
> inside the enterprise firewall. 

I think we're encountering situations where programmers who have been
trained very well to work in those situations now want to take the same
approaches and make them work over much wider distances.  

I suspect that API-based approaches have natural functional limitations,
which may not be encountered in tightly governed and deliberately
homogenous systems (or very simple systems).  APIs offer a clear set of
rules, but at a distinct cost to those for whom the design of the API is
suboptimal.  

(There's a bit of a parallel to relational databases, which are great
things unless your data doesn't mesh well with tables.)

Loosely governed messaging may be, in some senses, sub-optimal for
everyone, but the flexibility there offers an opportunity to optimize
exchanges as necessary.

> For the autonomous nodes of the internetwork topology, the capacity to 
> execute transactions--and realize profit--depends on processes which are 
> sufficiently versatile to do something useful with the wide range of data, 
> in often unanticipated forms, which might become available.

I suspect the same possibilities are inherent in most systems, even
those inside the firewall. The frameworks we've built those systems on
tend to snuff out the lines of thinking which might have brought us
there.