What are the differences between CPAN and, for instance, Maven repositories?
Maven repositories are just that: Repositories. They store file and display some metadata.
CPAN shows documentation from the code of a distribution itself. It also provides a bugtracker. When installing things from CPAN via cpan on the command line, the tests of that dist are run. Users can opt to install Test::Reporter, which then uploads those test results to cpantesters.org. Many people do this, across many OSes and Perl versions across the globe. Authors of dists get emails of failures. Users can look at nice matrices displaying the passes and failures on their platform combination. On failures regression analysis is performed automatically to try and pinpoint possible failure reasons. Since metacpan there is also a public API to an ElasticSearch database containing all the metadata about CPAN dists and their authors. So you can for example, find all dists written by authors living in your town.
Et cetera.
I could go on, but i think i made my point.
I want this package installed (plus its dependencies), and I want it now, that's all.
I'm not sure I buy that (paraphrased) "Perl is good because CPAN has billions more packages than other language's repos" either, for one I remember trying to find a library that did SHA1 digests, there were three, and I managed to pick the one that give incorrect results...
of course in Python it's just built-in, as are sockets, http, smtp, imap...
For one, it means the dists are of higher quality because authors have better tools at their hands and get poked to increase quality.
For the other, as a user you can tell whether the dist works on your particular system without even installing it and if it doesn't, you can easily find out on what system it DOES work, or which previous version did work on your system: http://www.cpantesters.org/distro/M/Moose.html#Moose-2.0205
> I managed to pick the one that give incorrect results...
That's why there is a review system:
http://cpanratings.perl.org/dist/Email-Send http://cpanratings.perl.org/dist/Email-Sender
I guess when you have a language that doesn't have a specification, you require things like this...
That is why i said the other languages have a long way to go yet; no matter how much you try to shift goal posts or resort to out-of-band attacks.
the review system is a questionable advantage too, as I don't want to go digging through 16 SMTP packages to find the best one, and especially as I suspect most people install CPAN modules through the command prompt (or via their distro's package manager), rather than digging through a website.
I also note that Python tends to adopt the best packages into the standard library...
Jeeze, you're barely making any sense. I have to wonder if you're trolling here. My point was about the utility of the ecosystem, which is useful for any language. The dists i mentioned are just examples of distributions you could find on CPAN other than Moose which all receive the same treatment.
> I suspect most people install CPAN modules through the command prompt (or via their distro's package manager), rather than digging through a website.
The installation is done via command line, but it is always preceded by research on the websites. Would you exclusively buy movies by title alone without looking at reviews?
> I also note that Python tends to adopt the best packages into the standard library...
Perl core developers do include useful things as well, but they are fairly strict about which ones, since they prefer to leave them on CPAN so they can be developed and released at a faster pace than the core. And in fact, the core is increasingly being viewed merely as a tool to access CPAN.
It doesn't matter how competant your product is (CPAN) if you aren't delivering what the market wants.
Yes, CPAN provides a utility in the ecosystem, but most users do not use the extra features that CPAN provides over something like PyPi.
And even in the cases where there is a feature overlap (e.g. repository storage and breath), after a certain point things are "good enough" and theres a marginal return for adding more.
Does it matter if there are 17 SHA1 implementations on CPAN and 3 on PyPi? Not really, since you only need 1 good implementation. In fact it could be argued that having more implementations is worse due to paradox of choice.
How do you know that?
Let me add also mine anecdotal experience from yesterday:
I looked for a good SQL parser/evaluator in Python (which I partly use at work). There might be one, but I didn't find any. It is easy to find on CPAN. (And this is hardly an unusual request.)