[% setvar title More modules %]
|Note: these documents may be out of date. Do not use as reference!|
To see what is currently happening visit http://www.perl6.org/
Maintainer: Michael G Schwern <firstname.lastname@example.org> Date: 19 Sep 2000 Mailing List: email@example.com Number: 260 Version: 1 Status: Developing
Perl should come distributed with more modules.
Currently, in order to get a really useful Perl installation you have to immediately hit CPAN and download a battery of modules. This is kind of silly. Its a tedious step that can be eliminated if we expand what Perl comes with.
When a "how do I do X" type question pops up and is answered "module Y does X", a typical response is "is module Y in the core?" There is some trepidation about using CPAN modules because they must be downloaded and installed (often a difficult problem for programs which are to be distributed, especially to Windows and MacOS), and because of perceived instability.
Core modules also establish a lowest common version. If version 1.5 of module Y is distributed with Perl, yet version 1.6 is on CPAN, a program author can code for 1.5's API and be fairly certain it will work on most installations.
The most obvious objection is core bloat. 5.6.0 is already over 5 megs and only going to get fatter. Throwing lots of modules into the core will only make it even bigger, possibly getting up to 10 megs! There are a few solutions for this.
The first is to provide several distributions of Perl. A minimalistic distribution might provide just perl and a handful of modules. Another provides the docs. Another a set of modules equivalent to the current set and yet another provides our extended set. CPAN/src/perl-X.X.X.tar.gz would be one big bundle encompassing all these sets. Other combinations might be made available.
Another is to provide several different install options. "make install" installs everything, as usual. "make small_install" installs just the current set of modules. "make tiny_install" installs a bare minimum, not even docs.
There's alot of crufy stuff in the source distribution that could be dumped. Eliminating lib/*.pl (except perl5db.pl) and eg/ will free up over 200k which makes room for a couple more modules.
Finally, we've got about 18 months before Perl 6 is expected to ship. In that time, DSL and cable modems will become more and more ubiquitous (my dad has one!) and pulling down 10 megs will be less of an issue for Joe Average. Users still on low bandwidth can get one of the slimmer distributions.
Some of the more obvious modules are:
Perl talks to databases alot, and DBI is how you talk to databases with Perl. Distributing it would push Perl even further as a powerful database access language. DBD::CSV and DBD::RAM (along with their supporting modules) would also go along, so even without a database DBI is still useful for prototyping and provides a path for programs to migrate to a real database when one is needed.
DBD::ODBC would also be nice, as most databases speak it. The free ODBC drivers should also be distributed. They are firming up now and by the time Perl 6 is out should be stable and well supported.
Rumor has it Perl has been used for some web stuff. LWP is the most obvious module for this sort of work.
LWP is quite fat, perhaps the maintainers could be petitioned to split it up into a basic set and a set of extensions. Perl would distribute the basic set.
This Internet thing is getting pretty popular, and Perl should come ready for it. libnet provides all the basic network clients. FTP, NNTP, POP3, SMTP...
"How do I send mail?" More often than not this question winds up with
open MAIL "|/usr/lib/sendmail -o". This is not only
non-portable, its also ripe for security holes. Mail::Send, or
similar module, makes sending mail easy, safe and portable.
The exact set is not so important as a consensus that its a Good Idea to start distributing more modules.
The integration and upkeep of existing CPAN modules is already a fairly well-understood procedure. The only foreseeable problem is configuring and testing those modules requiring network access, as this is not always available.
RFC 228 - Add memoize into the standard library