Computer Language Popularity Trend

[cartercc at gmail.com]

I, too, attempt to track the popularity of computer languages, but I
like to look at the job boards. My theory is that the number of
employers looking for particular skills indicates the relative
popularity of the language. This is a somewhat crude measure,
particularly with Microsoft technologies (VB, VB6, VB.NET, VS, etc). I
think it's much more reliable with open source languages, such as Java,
Perl, PHP, and so on.

'Popularity' is a slippery concept as well. C isn't real popular in
terms of jobs, but it is in terms of compensation. In system
administration (which I also follow), Windows has large numbers of
jobs, but a low level of compensation. OSes like AIX on the other hand
have lower numbers of available jobs, but those tend to be more highly
compensated. One could argue that compensation is a function of
popularity, with the more unpoular technologies having carrying a
bigger price to attract more people -- an example of supply and demand
-- but then one would have to argue that garbade collectors should be
more highly compensated that physicians.

You can also get a rough measure ot the popularity of web scripting
languages from an analysis of the URLs. The last time I did this was in
2003, and as I recall, these were the results:
PHP 30% and increasing
Perl 28% and falling
ASP 25% and falling fast
ColdFusion 6% and steady
Java and JSP 5% and increasing
others, Python, Ruby, ...

Again, this is a very rough measure. Java, for instance, is used by big
companies (like auto manufacturers, aerospace industries, defense
contractors, big retailers, etc.) One site/one vote isn't
representative necessarily, plus the bigger companies employ more
people than the smaller companies that tend to use FOSS.

Finally, in my area, we have a lot of banking and insurance jobs. These
companies internally are exclusively Microsoft shops. It's virtually
impossible to work there unless you know Visual Studio and SQL Server.
Misrosoft people tend not to prowl the newsgroups, and I would suspect
that any measurement based on numbers of newsgroup postings would be
skewed for this reason.

CC