The real question is which modules can simply be 'installed' by copying over the .pm files, and which require actually being compiled on the machine in question.

Some modules use AutoSplit, and so have .al files, in addition to the .pm files. Some modules have XS components and need to be built with a C compiler, and the binary .so (or whatever) files from one machine may not be compatible with another machine. There are other factors that could make it necessary to build and install the module the regular way.

Fortunately, it's ridiculously simple to install a module, in the standard way, into any directory you want:

perl Makefile.PL PREFIX=/path/to/my/lib/directory
make
make test
make install
[download]

You could even set up a CGI script to do it!

I do this sort of thing all of the time on FTP-only ISPs. I have a RIDICULOUSLY simple barebones template of a CGI script which I modify to run arbitrary commands and display their output.

#!/usr/bin/perl -w

use strict;
use CGI qw(:all);
use CGI::Carp qw( fatalsToBrowser );

my $v = `arbitrary command 2>&1`;

print join( "\n",
            header,
            start_html("Arbitrary Command"),
            strong(h1("Arbitrary Command:")),
            pre( $v ),
            end_html ),"\n";
[download]

Just customize 'arbitrary command' and save the result as, say, 'arbitrary command.cgi' using FTP and Bob's your uncle.

dmm

You can give a man a fish and feed him for a day ...
Or, you can teach him to fish and feed him for a lifetime

That won't, sadly, answer this monk's enquiry, since said monk specifically mentioned that he is limited to FTP access. I ran into this myself just the other day (:twitch: CGI.pm is at version 2.36 there).

On the other hand, the idea of writing a CGI script to do it, could work... insane, twisted, wrong, but it could work.

I noticed that the monk said only FTP access was allowed. That is specifically why I said it could be done through a CGI script. Here's an example of what I meant.

#!/usr/local/bin/perl

use CGI;
my $cgi = new CGI;

print $cgi->header('text/plain');

my $source_dir = "/home/me/source";
my $lib_dir    = "/home/me/lib";

my $module = $cgi->param('module');

chdir($source_dir) or die "Can't chdir to $source_dir: $!\n";

system("/bin/gunzip $module.tar.gz") and die "Can't gunzip $module.tar
+.gz\n";
system("/bin/tar xf $module.tar") and die "Can't untar $module.tar\n";

chdir("$source_dir/$module") or die "Can't chdir to $source_dir/$modul
+e: $!\n";

system("/usr/local/bin/perl Makefile.PL PREFIX=$lib_dir")
  and die "Error running Makefile.PL\n";
system("/usr/bin/make") and die "Error executing make\n";
system("/usr/bin/make test") and die "Error executing make test\n";
system("/usr/bin/make install") and die "Error executing make install.
+\n";
[download]

So, all one has to do is upload this script, upload the .tar.gz file to the source directory, and hit the script from the browser with the name of the module in the module parameter. All possible simply with FTP and CGI.

Obviously, the script needs better error checking and reporting, and it should capture STDERR from the system calls and send it to the browser along with STDOUT. (Calling a single shell script to handle all the unpacking and building might be an improvement.) I've just written this as a quick hack to get the general idea across.

Update: Please see DrManhattan's important addendum below; using this script as is would be a huge security risk! An improved script would use taint-checking and restrict the module name to a limited set of characters.

...the @INC array is globally prepended by your 'use lib' line, right Monks?

I honestly don't know, but otherwise the following should do it:

BEGIN{unshift(@INC,'.');}

f--k the world!!!!
/dev/world has reached maximal mount count, check forced.

Modules like Image::Magick (which is just an interface to some C programs, image libraries, and whatnot) will get *quite* interesting.

I would imagine that the C libs wouldn't be there, and that you'd have to build imagemagick with

./configure --prefix=$HOME
make
make install
[download]

Then of course, the system doesn't know where to find your libs...

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${HOME}/lib
ldconfig
[download]

Then you'd need to build the Perl interface...

cd /path/to/Image_Magick
perl Makefile.pl PREFIX=${HOME}/lib/perl5
make
make install
[download]

Of course then when the webserver (probably user nobody) tries to run it... might not find the libs... depending...

So the CGI might start:

#!/usr/bin/perl -t
use strict;
use lib '/home/myhome/lib/perl5';
$ENV{LD_LIBRARY_PATH} = $ENV{LD_LIBRARY_PATH}.'/home/myhome/lib';
use Image::Magick;
[download]

Yeah, that might be... "interesting."
--
Snazzy tagline here

Thanks :)

So is there hope that HTML::Mason could work even though it requires yet even more packages that are not on the system by default?

My understanding is that you would have to install all of those Modules that it uses into your local lib directory specified in your use lib statement. This is only for modules that are not already in the normal @INC path.

I used to run on eHost (let's not go there) and it was the same kind of FTP and no ssh or telnet type <stuff>. Very limiting. As to faking command line access through a script: Most likely any cgi script you write that opens a system pipe and tries to run commands will fail because the script will most likely run as either your user or as 'nobody'.