Re: Just for the record

From: Bob Badour <bbadour_at_pei.sympatico.ca>
Date: Wed, 28 Jun 2006 20:42:28 GMT
Message-ID: <UQBog.3417$pu3.83055_at_ursa-nb00s0.nbnet.nb.ca>


kvnkrkptrck_at_gmail.com wrote:

> Is my understanding correct?
>
> Dijkstra: SOC = When solving a complex problem, focus on one aspect of
> the problem (correctness, efficiency, etc.) while bearing in mind that
> the other aspects exist.

Yes, that's an excellent description. He even goes as far as separating a concern for relevance or need from a concern for correctness. ie. One should always start with: "Should I even be bothered with this?" before proceeding to: "Is this correct?" and "Is this efficient?"

> Kiczales: SOC = After the vertical OO and horizontal AO decomposition
> of a complex system, handle resulting modules separately.

I wouldn't even call them modules. Perhaps something more like functional requirements or artifacts. The AOP use of 'concern' is simply a muddled mess. The whole method of AOP strikes me as seriously flawed as it undermines the concern for correctness.

The frequent AOP habit of using logging as an prime example even strikes me as weak. Reading about Dijkstra's initial alarm at the prospect of adding real-time interrupts to a computer reminds one that, with multiple threads of execution, even program traces are useless for demonstrating correctness. Better formalisms and abstractions are required -- not implementation hacks. Received on Wed Jun 28 2006 - 22:42:28 CEST

Original text of this message