Re: Programming language popularity

[Gautier write-only <gautier_niouzes@hotmail.com>]

On Apr 3, 11:46 am, "Martin Krischik" <krisc...@users.sourceforge.net>
wrote:

> Tiobe sells services for curly braced languages.  So I am not all that  
> convinced about Tiobe - especialy after they removed usenet in favour of  
> youtube. I can understand adding an addition source - but removing an  
> existing source for which all the scripts are already written. That sounds  
> like fine tuning the statistic so the right result is reported.

It doesn't explain everything - as you notice later...
For instance usenet is partially mirrored by sites which in their turn
appear in standard web searches. Nevertheless I plan to introduce
groups.google.com into the LPI as it reflects not only the whole
usenet but also many popular alternative forums.

> At comp.lang.ada we recently did our own stat which too comes to different  
> result: http://lang-index.sourceforge.net using similar tests as Tiobe.

I need to stress there that the LPI's results can be verified and
reproduced, and that all intermediary results are published.
Now where the differences comes from, since currently the methods are
very close ?
Apart from slight changes in confidence factors, perhaps the main
reason relies in the way tiobe's data are collected, which seems to be
very time-consuming (cf the FAQ). Perhaps there are typos: for
instance, Fortran made a jump from some 0.35% to 0.563% in March (with
a nice headline about good old Fortran being back in the top 20 !) and
the moved back to 0.35%. Although such movement may well be caused by
some issue in one search engine or the other... There seem also to be
a accumulation of biases. Perhaps something else...
Anyway, I would not buy the data without a full disclosure or an
audit.

My 0.02 CHF

G.