Stories
Slash Boxes
Comments
NOTE: use Perl; is on undef hiatus. You can read content, but you can't post it. More info will be forthcoming forthcomingly.

All the Perl that's Practical to Extract and Report

use Perl Log In

Log In

[ Create a new account ]

Ovid (2709)

Ovid
  (email not shown publicly)
http://publius-ovidius.livejournal.com/
AOL IM: ovidperl (Add Buddy, Send Message)

Stuff with the Perl Foundation. A couple of patches in the Perl core. A few CPAN modules. That about sums it up.

Journal of Ovid (2709)

Friday July 04, 2008
04:56 AM

Rethinking the CPAN Problem

[ #36852 ]

Recently on Perl-QA there's been a debate both rather lengthy (going on for well over a week) and a touch acrimonious. It's actually been rather civilized and the acrimony has been more in the form of blunt comments, but it boils down to people still complaining about the "CPAN" problem:

The CPAN is huge and hard to evaluate

They weren't really talking about this problem. They were talking about tools surrounding this. In particular, there were sharp disagreements about the utility CPANTs and many people pointing out how things can be taken out of context. For example, both Damian Conway and Mark Jason Dominus have relatively low CPANTs scores and this certainly says something about their distributions. If you're experienced in the Perl community or new to it, you could draw drastically different conclusions as to what this "something" is.

Still, when you look at AnnoCPAN, cpanratings, CPAN RT, CPANTs, CPAN testers, the module itself, its competition, documentation, etc., you can quickly get overloaded with the bewildering array of choices (Class:: namespace, anyone?). I think, though, that these attempts to provide more information are good and people are misunderstanding the core problem.

Many people complain that the CPAN is huge and hard to evaluate, but that's not the problem. The fact that the CPAN is in the state it is in is largely because it's reflecting the real world. Let's try a little experiment: if you know nothing about content management systems, but you want the "best" one to run use.perl and your boss tells you it must not be slashcode, what do you choose? I'll wait. Tell me how long it takes you to choose the "best".

"But wait!", you protest. "What does 'best' mean?" Is it cost? Is it programming language? Is it ease of use once learnt? Is it easy of learning? Is it ease of setup? How much maintenance will it entail? Who else uses it? How long has it been around? What's its history vis-a-vis security issues? Does it rely on technologies that our IT department won't support? Do we have to change anything internally to use it? And the list goes on and on and on ...

You see, in the real world, when you choose software, whether you've written it yourself or not, you have a set of requirements and you have to evaluate the software against your needs. This is often very hard. CPAN, thus, mimics the real world to a certain extent.

There is an interesting difference, though. As a general rule, CPAN authors and those creating collaborating informational sites are interested in providing you with real information, not marketing spin. This can be a huge win, with the caveat that the author may not know what real information is needed.

What we really need is to incorporate one-stop shopping for these various resources to give us an ability to evaluate them in context. CPAN distributions need "tags" which people should be able to upvote or downvote based on appropriateness. Thus, even though "File::Find::Rule::XPath" might get returned as a search for "xpath", if you narrow your search with "xml", it would lower the likelihood of getting FFRX because even if it's been mistagged with "xml", people would downvote said tag and thus reduce its weight.

With that, we need solid APIs for the other services so that the consumer, once presented with a set of choices, can consider how well a module is maintained, last release date, bug count, annotations, etc. Which information is important? How the heck can I know? It's often subjective, anyway, and you and I could reasonably, disagree. Give 'em all the information we easily can, regardless of our personal opinion of its worth.

CPAN itself should probably not be the best place for this (though by default "search.cpan.org" is a poor man's version of what I envision). Instead, it should be the canonical repository and a front-end tool which allows intelligent searching, filtering, list of relevant articles and aggregation of all other relevant data that we can think of. It should merge the available information in one spot to ease the burden on the poor author.

This can be done, but whether it will be done is another story. I think, however, that this is what we need. I don't want more restrictions on the CPAN. I don't want to tell people developing external informational tools how they should do their stuff. I want one-stop shopping for most of this information and let me decide what's important. This would ease one of the biggest burdens we have.

The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
 Full
 Abbreviated
 Hidden
More | Login | Reply
Loading... please wait.
  • Gabor has put some effort into adding tags to CPAN with CPAN::Forum [cpanforum.com]. There even is a tag cloud [cpanforum.com].
  • Sounds like we really need the CPAN Metabase [perl.org]... (scroll down to 'Project 2 -- CPAN Metabase for CPAN Testers 2.0')

  • I strongly agree, tags are the way to do this -- they're great for this kind of "needle in a haystack" searching.

    Another thing: mid-length module descriptions, of a roughly freshmeat-style length (512 chars). this is short enough to fit in one paragraph, and therefore fit loads of modules on a single page, but allows enough text to specify function and purpose with a few synonyms for searchability.

  • For the past three months, we've had a mailing list, rethinking-cpan, for discussing this very problem. Here's the blog post that started it: http://perlbuzz.com/2008/04/rethinking-the-interface-to-cpan.html [perlbuzz.com]

    I've started a group, rethinking-cpan [google.com], for discussing the ideas I've posted here. -- Andy

    Every few months, someone comes up with a modest proposal to improve CPAN and its public face. Usually it'll be about "how to make CPAN easier to search". It may be about adding reviews to search.cpan.org, or

    --

    --
    xoa

  • Years ago, I convened that CPAN meeting in London for the sole purpose of getting some sort of Metadata standard together as this is an old problem that could possibly be partially solved with metadata. Unfortunately, not much happened in that vein as it would require a lot of work and cooperation from more than a few people and consensus has never been a strong point of perl people. Andreas, Ken, Leon and Jos were the most interested and may possibly be still....if you're interested in picking up that idea
  • With webapps, the site maintainer has to be bothered to add every new thing. Further, the overhead to setting-up a local copy of a webapp often involves install and setup of a particular database, web server, etc. Webapps (by nature) don't need to be easily deployed onto a wide variety of systems, so setup usually kills any ambition of trying to add "one simple idea" to a thing -- even if you have the source code.

    What if you could download and install a CPAN search+install application on your computer in