I've never had much success with searching CPAN by functionality - often I want something specific which may not appear in the module name or description (I don't know if search.cpan.org even looks at descriptions).

A better approach, I've found, is to look for previously asked questions by searching for likely phrases on various sites (Perlmonks being foremost). Example: I want a module that does X, search for "how do I do X", "problem doing X", "doing X doesn't work" (more or less, allowing for proper grammar etc. depending on the value of X).

This usually shows up code with use statements or module recommendations or (best case) admonitions to "just use Some::Mod, you'll want to eventually anyway" (WWW::Mechanize for most web client stuff would be an example of the latter).

how do I know which is best to use for simple/complex logging on files?

As for your question, what is "best" depends on what you think it good in a module. Do you rank speed over ease of use? Do you prefer objects over a procedural interface? And when is logging simple or complex anyway? Considering those answers will vary from person to person, it would be very hard for someone to say "you should module X" - assuming you want that someone to have your best interest in mind. Not to mention said someone should reevaluate which module is best each time a new logging module hits CPAN.

My approach so far is keeping up with knowledge and use by experience. After that I use Perlmonks search. When I don't find it I usually ask here. A better approach would see me using modules in better ways.

It may not be as straight forward as sf.net. In Perl, there is more than one way to do it. It is for the community but also by the community. It is not about use one and forget the rest. It is about contributing in such a manner that others can directly benefit from it. The few sections on Perl such as Q&A and Snippet provides this nicely.

Said that you can try searching here http://cpanratings.perl.org/ for previously reviewed modules. Tagging CPAN modules by community would help enhancing system further. We can have also 'if you are using this module', you may also want to use that module, where each user can put module suggestion as they work along. We have to configure it to use with practice, rather than knowledge dump.

Just ask. Here.

Or search (here), as someone else has probably already asked.

"why can't CPAN give such a helpful search results" ... because it can't.

1. Relevance - looks to me like this is how it's sorted. The relevance according to your search against the module names and descriptions.
2. Activity - activity may mean something useful for big projects, where there's (almost) always what to add, what to improve, what to fix. This is not the case with small libraries. There are often used, recommended modules, that had not been updated in months. Sometimes years. Because there's no need to add anything, nothing to fix and the author not the users feel the need to improve what works well and is tested and proven.
3. Rank - please define.
4. Registered - you mean the date? I don't see how would it help.
5. Latest file - see above about Activity.
6. Downloads - no we can't give that. Keep in mind that CPAN is not a centralized resource. It's comprised of many mirrors ... you can't get the download statistics for all. Plus you would have to add the stats from PPM repositories. Otherwise the numbers especially for Windows related modules would be almost zero.

With all this said ... what would you base your Rank on? There are some reviews and some ratings, but there are far too few to be dependable.