Re: [Algorithm] Sum of Primes < 1000000

Lew wrote:

The optimization there was of a database
structure, where the difference for one report was between O(n^2) and
O(n). I was forbidden to make the algorithmic optimization, then got in
trouble because the report was too slow once they loaded all the data. It
took months and lawyers to straighten that one out.

Chris Uppal wrote:
I think part of the problem with this sort of situation is that we don't have a
commonly understood way of adding contingent sub-plans to projects. It seems
quite plausible to me that your manager might have correctly decided that the
risk and time involved in the optimisation was not acceptable at that time.

The fix involved half a day's work and was routine - no risk.

But ideally you and s/he should have been able to add a contingent extension to
the project plan, somewhat like (briefly)

Risk:
The XYZ report may run too slowly with production-scale datasets.

There was no doubt that it would, given the projected data sizes.

Reason:
O(N^2) algorithm used (unecessarily).
Responsibility for monitoring and reviewing risk:
<some person, group, or organisation> at <some time>.
Suggested fix:
<details>
Probable time:
<N weeks> + testing.
NB, retesting required /at system integration scope/.

No such reasoned approach was forthcoming from my project manager then. He simply didn't want to hear it. The difference in that case was not weeks but a few hours of development time to implement the fix.

The "Responsibility for monitoring and reviewing risk:" bit would be, to my
mind, the single most important bit of this in that the system couldn't be
deemed complete unless the issue had been resolved.

I had a fix ready to implement. I made sure my replacement knew of it. When he in turn was finally permitted to make the change, it took half a day and completely fixed the problem. That didn't happen until the project manager who had forbidden the change was replaced. By the guy I'd trained with the fix.

At least, /I/ have never seen anything like this used in project planning (not
even informally).

Not sure of the antecedent. If you mean you've never seen someone irrationally and arbitrarily refuse to let a product work when it would have been easy to repair, lucky you. If you mean you've never seen anyone plan for such a fix, using reasoned analysis and with fallback positions, unlucky you. I have seen both ends of that spectrum.

- Lew
.

Quantcast