How Fresh is the CPAN?

by chromatic

I like numbers. They can mean a lot of things.

Rather than continuing silly arguments over obfuscated and flawed measurements of "language popularity", perhaps a better way of measuring the viability of a language or platform is to measure the freshness of its ecosystem.

LaPerla's How Fresh is the CPAN? measures the upload dates of one of the world's largest and most active repositories of free software. Of the 12,000 (or is it 14,000 now?) distributions on the CPAN, 25% have a most recent upload date of February 2008 or newer. Half have an upload date of 2007 or newer.

You don't get those kinds of statistics by putting "Ruby Programming" into Google and pretending the results are meaningful.


Sean Blakey
2008-05-08 20:43:29
So, we have a chart expressing CPAN's freshness - effectively, a single data point. Can this be compared to archives for other languages? Do we have any idea how the "freshness" of CPAN has been changing over time?

Without comparison, there is no meaning. Sure, CPAN is still growing more-or-less exponentially, but how does that contribute to any sort of "opinion"?

Noah Gift
2008-05-09 06:28:57
Python's Package Repository is growing by leaps and bounds. In fact, in our book, we have a whole chapter dedicated to teaching people how to get their packages into the Python Package Repository. With any luck the number of Python packages, libraries, scripts, will grow exponentially.
Aaron Trevena
2008-05-09 11:30:01

It shows several things - CPAN is growing, the number of users using and contributing to CPAN is growing, you can combine this with perl job market and user group data and show that perl usage is growing.

Unlike TIOBE which is basically googlefight relabelled as usage/market metrics (and still get's it wrong - nobody has been able to replicate tiobes results with the same data and/or claimed techniques)

2008-05-12 04:16:57
These stats don't prove Perl is better than language X. They don't even prove that Perl hasn't declined in the last few years. But they do prove that there is a fair bit of Perl activity around. It's not dead, and probably not even dying, which seem to be popular memes.
2008-05-13 07:47:22
I think all numbers are good in their own way, even the TIOBE index. I think they (or everyone linking to it) simply go wrong in assigning more value to their metric than it actually has.

Here's my site for looking at language popularity: My idea is that all "public facing" (i.e. Internet) metrics have some deficiency or another, and that (1) the more you can show, the better your perspective, and (2) transparency of the data collection process is good -- e.g. I let you download and inspect the code that creates my metrics.

Also, I agree that package repository metrics would be very valuable. In fact, I have a couple similar metrics, one for Ohloh (not really a repository, but aggregates code source from them), and SourceForge. Adding language specific package repositories is one that's next on my list.

Andy Lester
2008-05-15 16:24:16
Without comparison, there is no meaning.

Of course there is. The meaning is that CPAN is vibrant and well-updated. Whether it is more or less vibrant than any other has no bearing on it.

Vibrancy and importance are not zero-sum.

2008-05-16 21:36:28
Ruby doesn't need so many packages!