[RFC/P] Port System Re-Engineering

[Ade Lovett]

On Dec 03, 2007, at 09:42 , Aryeh M. Friedman wrote:
> If the questions are flawed then point out where.    If the general
> concept of a survey vs. user stories vs. what ever then state which
> you think is more productive.  If your problem is the medium/forum the
> data is being gathered in see below.

I'll say this one last time.

It is not a question of whether the questions are flawed, it is the  
communications medium in which they were posted.

You sent the "survey" to a single mailing list.  ports at FreeBSD.org

This is, statistically speaking, a self-selecting and exceptionally  
minor part of the community to which the survey is addressed, most  
notably:

1.  The survey will only be seen by people that currently use FreeBSD  
(subset 1)
2.  It will further only be seen by those that are aware of the ports at FreeBSD.org 
  mailing list and have taken the conscious decision to subscribe to  
it (subset 2)
3.  Within the readership of a relatively high volume mailing list (to  
wit, the aforementioned ports at FreeBSD.org), a significant minority, at  
best, will even notice the email (subset 3)
4.  Further, within that subset, you are soliciting responses, whereas  
folks in general are notoriously lazy, or simply believe that  
responding to such a random survey is a waste of their time and effort  
(subset 4)
5.  Taking the sub-sub-sub-subset of folks that use FreeBSD, are  
subscribed to the ports@ mailing list, have read your survey, and  
responded to it, you then have to do analysis on what percentage of  
responses are incomplete, flamebait, inaccurate, or otherwise flawed.

For example, compare the number of discrete folks that are, to one  
extent or another, contributing to this thread, as a percentage of  
total FreeBSD installed systems.  Care to guess what that percentage  
is?  Unless it's significant, any results can be massaged into  
anything at all.  Without a statistically valid sampling, it is simply  
not possible to make any conclusions with whatever dataset exists as a  
result of said survey.

Stats 101.

-aDe