On 09/14/2011 04:27 PM, Andrei Alexandrescu wrote:
Words like "demonstrated" or "clarified" are a bit exaggerated within
this discussion, as is "flawed" when describing what Tiobe does.
Only if you are delusional and can't recognize the fundamental problem
with search results getting worse and worse the farther out you go (who
hasn't experience this with their everyday searches?), and then only
looking at the first 100 results to get a simple percentage for the
whole distribution. On top of this, D matches all kinds of stuff,
whereas a name like Scala does not.
I also gave some simple and hard stats that agreed with each other about
just how lopsided the results for Scala were for D, yet D is at Tiobe
#20, and Scala is at #50.
Not to mention the implication that I lack integrity, which is quite
a bit out of character even by the standard set by the corrosiveness
of your past posts. Is defending your view of Tiobe worth all this?
The way you keep on defending Tiobe isn't helping. There have also been
other instances in the past where I've found less than savory
statements, and I've already mentioned these before.
The statistics I quoted were varied, simple, and easy to verify.
They're not ground truth; for example, it's known there is a lot of
noise in Google's estimate number of pages.
And the Amazon sales rank? And the number of books available? And the
number of questions on Stack Overflow? Should I start looking for number
of jobs next? I'm willing to be you lose there, too.
For example, searching for the exact phrase "scala programming"
yields 149,000 estimated results, whereas searching for the exact
phrase "scala programming language" yields 418,000 results. This is
incorrect because the latter query includes the former. To some
consolation, the relationship between the estimated result set sizes
for e.g. "d programming" and "d programming language" looks more
plausible (343,000 vs. 199,000).
What you've done is demonstrated that Tiobe's method is even more
flawed. You need to sample the distribution to get valid results, not
just the first 100 pages.
