Comments on Carlo Pescio - blog: Design, Structure, and Decisions

thought-provoking a post as ever! -- Your r...

2010-11-18T14:02:48.867+01:00

thought-provoking a post as ever!
--
Your reply too! :-)

Looking at your (very) insightful ideas from the traditional -ilities perspective, I could say that they translate rather well to predictability.

For instance, on the run-time side, all the multiple-cache levels, branch prediction, speculative execution, etc that have been introduced at the hardware level make "average" performance good, but also make performance prediction much harder. Conversely, the first generation of Cray supercomputers didn't have any cache at all. That made predicting performance much easier, which was good, because back then the real issue was making parallel software working at all.

On the artifact side, you're right to say that we can keep a design safe or go to a cliff, and I would add (quoting Tom Peters, I think) that sometimes it's not safe to be safe. Some problems call for what Walter Vincenti defined as "radical design".

Incorporating those concepts in the Physics of Software is not trivial. It's rather easy to say that some decisions open the door to a different, unexplored forcefield. Those are pivotal moments in the life of a project, where choices have a nonlinear impact. It's something entirely different to translate that into something more formal. I'll have to think hard about it :-).

As usual, there is also a human side of the story. Some people just can't stand the discomfort of going out on a limb and explore uncharted territories. I've experienced that first-hand, trying to push some teams away from a mediocre (or plain bad) but "safe" design and toward something with a much higher potential, but also more challenging and way outside their comfort zone. Dealing with resistance was much, much harder than making the right design choices :-)

Dear Carlo, thought-provoking a post as ever! I&#...

2010-11-15T18:54:41.717+01:00

Dear Carlo,

thought-provoking a post as ever!
I'd like to discuss more deeply a particular point: "the need to make a decision".

The fact is, our "needs" are of two kinds: those which are inflicted on us by reality, and those which we can see in advance through the "telescope" of our knowledge.

What I mean is: every choice we make can dramatically enlighten or cloud the path we take from there.

Allow me to borrow a metaphor from the clinical practice.
Let's say a patient comes to me at the first signs of a neurodegenerative disease.
I choose a drug balancing different forces: efficacy, side effects, patient's compliance, etc etc.
Now I come to the point: let's say I choose quite a powerful drug. So what?
Well, the next time I see my patient, I'd better question him in deep about further signs which pehabs he did not notice BECAUSE OF the efficacy of the drug! The drug has lightened his symptoms, clouding the progress of the disease.
Was it a legitimate choice? Absolutely. But I must be aware that there is a trade-off between the solution to the actual problem and the predictability of the problems to come.

I followed with great attention your writing about the "physics of software".
I would like you to take into consideration an additional factor: what we could name "the horizon-effect" of a choice or an artifact.
Some design are perfectly sane pathways running at the bottom of mountain. Some others are harder, but put us at a peak, enabling the designer to see in advance the obstacles ahead.

Yours,
Guido Marongiu
guido.marongiu@gmail.com

2010-11-03T15:43:53.287+01:00

This comment has been removed by a blog administrator.