summary refs log tree commit diff stats
path: root/plugins/perl/lib/Pod/Html.pm
diff options
context:
space:
mode:
authorTingPing <tingping@tingping.se>2014-01-20 01:14:10 -0500
committerTingPing <tingping@tingping.se>2014-01-20 01:14:10 -0500
commit681e14b3b823c3e700812b19c1f561e6370df0c4 (patch)
tree03034d829ad0b945bb2203de7383cba0cfa0ff8a /plugins/perl/lib/Pod/Html.pm
parent38cbabea026fe47d84da156772f5a9a97bd401f5 (diff)
Remove unused perl files
- old example scripts
- outdated docs
- mingw build script
Diffstat (limited to 'plugins/perl/lib/Pod/Html.pm')
-rw-r--r--plugins/perl/lib/Pod/Html.pm2399
1 files changed, 0 insertions, 2399 deletions
diff --git a/plugins/perl/lib/Pod/Html.pm b/plugins/perl/lib/Pod/Html.pm
deleted file mode 100644
index 3695564b..00000000
--- a/plugins/perl/lib/Pod/Html.pm
+++ /dev/null
@@ -1,2399 +0,0 @@
-package Pod::Html;
-use strict;
-require Exporter;
-
-use vars qw($VERSION @ISA @EXPORT @EXPORT_OK);
-$VERSION   = 1.08;
-@ISA       = qw(Exporter);
-@EXPORT    = qw(pod2html htmlify);
-@EXPORT_OK = qw(anchorify);
-
-use Carp;
-use Config;
-use Cwd;
-use File::Spec;
-use File::Spec::Unix;
-use Getopt::Long;
-
-use locale;    # make \w work right in non-ASCII lands
-
-=head1 NAME
-
-Pod::Html - module to convert pod files to HTML
-
-=head1 SYNOPSIS
-
-    use Pod::Html;
-    pod2html([options]);
-
-=head1 DESCRIPTION
-
-Converts files from pod format (see L<perlpod>) to HTML format.  It
-can automatically generate indexes and cross-references, and it keeps
-a cache of things it knows how to cross-reference.
-
-=head1 FUNCTIONS
-
-=head2 pod2html
-
-    pod2html("pod2html",
-             "--podpath=lib:ext:pod:vms",
-             "--podroot=/usr/src/perl",
-             "--htmlroot=/perl/nmanual",
-             "--libpods=perlfunc:perlguts:perlvar:perlrun:perlop",
-             "--recurse",
-             "--infile=foo.pod",
-             "--outfile=/perl/nmanual/foo.html");
-
-pod2html takes the following arguments:
-
-=over 4
-
-=item backlink
-
-    --backlink="Back to Top"
-
-Adds "Back to Top" links in front of every C<head1> heading (except for
-the first).  By default, no backlinks are generated.
-
-=item cachedir
-
-    --cachedir=name
-
-Creates the item and directory caches in the given directory.
-
-=item css
-
-    --css=stylesheet
-
-Specify the URL of a cascading style sheet.  Also disables all HTML/CSS
-C<style> attributes that are output by default (to avoid conflicts).
-
-=item flush
-
-    --flush
-
-Flushes the item and directory caches.
-
-=item header
-
-    --header
-    --noheader
-
-Creates header and footer blocks containing the text of the C<NAME>
-section.  By default, no headers are generated.
-
-=item help
-
-    --help
-
-Displays the usage message.
-
-=item hiddendirs
-
-    --hiddendirs
-    --nohiddendirs
-
-Include hidden directories in the search for POD's in podpath if recurse
-is set.
-The default is not to traverse any directory whose name begins with C<.>.
-See L</"podpath"> and L</"recurse">.
-
-[This option is for backward compatibility only.
-It's hard to imagine that one would usefully create a module with a
-name component beginning with C<.>.]
-
-=item htmldir
-
-    --htmldir=name
-
-Sets the directory in which the resulting HTML file is placed.  This
-is used to generate relative links to other files. Not passing this
-causes all links to be absolute, since this is the value that tells
-Pod::Html the root of the documentation tree.
-
-=item htmlroot
-
-    --htmlroot=name
-
-Sets the base URL for the HTML files.  When cross-references are made,
-the HTML root is prepended to the URL.
-
-=item index
-
-    --index
-    --noindex
-
-Generate an index at the top of the HTML file.  This is the default
-behaviour.
-
-=item infile
-
-    --infile=name
-
-Specify the pod file to convert.  Input is taken from STDIN if no
-infile is specified.
-
-=item libpods
-
-    --libpods=name:...:name
-
-List of page names (eg, "perlfunc") which contain linkable C<=item>s.
-
-=item netscape
-
-    --netscape
-    --nonetscape
-
-B<Deprecated>, has no effect. For backwards compatibility only.
-
-=item outfile
-
-    --outfile=name
-
-Specify the HTML file to create.  Output goes to STDOUT if no outfile
-is specified.
-
-=item podpath
-
-    --podpath=name:...:name
-
-Specify which subdirectories of the podroot contain pod files whose
-HTML converted forms can be linked to in cross references.
-
-=item podroot
-
-    --podroot=name
-
-Specify the base directory for finding library pods.
-
-=item quiet
-
-    --quiet
-    --noquiet
-
-Don't display I<mostly harmless> warning messages.  These messages
-will be displayed by default.  But this is not the same as C<verbose>
-mode.
-
-=item recurse
-
-    --recurse
-    --norecurse
-
-Recurse into subdirectories specified in podpath (default behaviour).
-
-=item title
-
-    --title=title
-
-Specify the title of the resulting HTML file.
-
-=item verbose
-
-    --verbose
-    --noverbose
-
-Display progress messages.  By default, they won't be displayed.
-
-=back
-
-=head2 htmlify
-
-    htmlify($heading);
-
-Converts a pod section specification to a suitable section specification
-for HTML. Note that we keep spaces and special characters except 
-C<", ?> (Netscape problem) and the hyphen (writer's problem...).
-
-=head2 anchorify
-
-    anchorify(@heading);
-
-Similar to C<htmlify()>, but turns non-alphanumerics into underscores.  Note
-that C<anchorify()> is not exported by default.
-
-=head1 ENVIRONMENT
-
-Uses C<$Config{pod2html}> to setup default options.
-
-=head1 AUTHOR
-
-Tom Christiansen, E<lt>tchrist@perl.comE<gt>.
-
-=head1 SEE ALSO
-
-L<perlpod>
-
-=head1 COPYRIGHT
-
-This program is distributed under the Artistic License.
-
-=cut
-
-my ($Cachedir);
-my ( $Dircache, $Itemcache );
-my @Begin_Stack;
-my @Libpods;
-my ( $Htmlroot, $Htmldir, $Htmlfile, $Htmlfileurl );
-my ( $Podfile, @Podpath, $Podroot );
-my $Css;
-
-my $Recurse;
-my $Quiet;
-my $HiddenDirs;
-my $Verbose;
-my $Doindex;
-
-my $Backlink;
-my ( $Listlevel, @Listend );
-my $After_Lpar;
-use vars qw($Ignore);    # need to localize it later.
-
-my ( %Items_Named, @Items_Seen );
-my ( $Title,       $Header );
-
-my $Top;
-my $Paragraph;
-
-my %Sections;
-
-# Caches
-my %Pages = ();          # associative array used to find the location
-                         #   of pages referenced by L<> links.
-my %Items = ();          # associative array used to find the location
-                         #   of =item directives referenced by C<> links
-
-my %Local_Items;
-my $Is83;
-
-my $Curdir = File::Spec->curdir;
-
-init_globals();
-
-sub init_globals {
-	$Cachedir = ".";     # The directory to which item and directory
-	                     # caches will be written.
-
-	$Dircache  = "pod2htmd.tmp";
-	$Itemcache = "pod2htmi.tmp";
-
-	@Begin_Stack = ();    # begin/end stack
-
-	@Libpods     = ();    # files to search for links from C<> directives
-	$Htmlroot    = "/";   # http-server base directory from which all
-	                      #   relative paths in $podpath stem.
-	$Htmldir     = "";    # The directory to which the html pages
-	                      # will (eventually) be written.
-	$Htmlfile    = "";    # write to stdout by default
-	$Htmlfileurl = "";    # The url that other files would use to
-	                      # refer to this file.  This is only used
-	                      # to make relative urls that point to
-	                      # other files.
-
-	$Podfile    = "";         # read from stdin by default
-	@Podpath    = ();         # list of directories containing library pods.
-	$Podroot    = $Curdir;    # filesystem base directory from which all
-	                          #   relative paths in $podpath stem.
-	$Css        = '';         # Cascading style sheet
-	$Recurse    = 1;          # recurse on subdirectories in $podpath.
-	$Quiet      = 0;          # not quiet by default
-	$Verbose    = 0;          # not verbose by default
-	$Doindex    = 1;          # non-zero if we should generate an index
-	$Backlink   = '';         # text for "back to top" links
-	$Listlevel  = 0;          # current list depth
-	@Listend    = ();         # the text to use to end the list.
-	$After_Lpar = 0;          # set to true after a par in an =item
-	$Ignore     = 1;          # whether or not to format text.  we don't
-	                          #   format text until we hit our first pod
-	                          #   directive.
-
-	@Items_Seen  = ();        # for multiples of the same item in perlfunc
-	%Items_Named = ();
-	$Header      = 0;         # produce block header/footer
-	$Title       = '';        # title to give the pod(s)
-	$Top         = 1;         # true if we are at the top of the doc.  used
-	                          #   to prevent the first <hr /> directive.
-	$Paragraph   = '';        # which paragraph we're processing (used
-	                          #   for error messages)
-	%Sections    = ();        # sections within this page
-
-	%Local_Items = ();
-	$Is83 = $^O eq 'dos';    # Is it an 8.3 filesystem?
-}
-
-#
-# clean_data: global clean-up of pod data
-#
-sub clean_data($) {
-	my ($dataref) = @_;
-	for my $i ( 0 .. $#{$dataref} ) {
-		${$dataref}[$i] =~ s/\s+\Z//;
-
-		# have a look for all-space lines
-		if ( ${$dataref}[$i] =~ /^\s+$/m and $dataref->[$i] !~ /^\s/ ) {
-			my @chunks = split( /^\s+$/m, ${$dataref}[$i] );
-			splice( @$dataref, $i, 1, @chunks );
-		}
-	}
-}
-
-sub pod2html {
-	local (@ARGV) = @_;
-	local ($/);
-	local $_;
-
-	init_globals();
-
-	$Is83 = 0 if ( defined(&Dos::UseLFN) && Dos::UseLFN() );
-
-	# cache of %Pages and %Items from last time we ran pod2html
-
-	#undef $opt_help if defined $opt_help;
-
-	# parse the command-line parameters
-	parse_command_line();
-
-	# escape the backlink argument (same goes for title but is done later...)
-	$Backlink = html_escape($Backlink) if defined $Backlink;
-
-	# set some variables to their default values if necessary
-	local *POD;
-	unless ( @ARGV && $ARGV[0] ) {
-		$Podfile = "-" unless $Podfile;    # stdin
-		open( POD, "<$Podfile" )
-			|| die "$0: cannot open $Podfile file for input: $!\n";
-	} else {
-		$Podfile = $ARGV[0];               # XXX: might be more filenames
-		*POD     = *ARGV;
-	}
-	$Htmlfile = "-" unless $Htmlfile;      # stdout
-	$Htmlroot = "" if $Htmlroot eq "/";    # so we don't get a //
-	$Htmldir =~ s#/\z##;                   # so we don't get a //
-	if (   $Htmlroot eq ''
-		&& defined($Htmldir)
-		&& $Htmldir ne ''
-		&& substr( $Htmlfile, 0, length($Htmldir) ) eq $Htmldir )
-	{
-
-		# Set the 'base' url for this file, so that we can use it
-		# as the location from which to calculate relative links
-		# to other files. If this is '', then absolute links will
-		# be used throughout.
-		$Htmlfileurl = "$Htmldir/" . substr( $Htmlfile, length($Htmldir) + 1 );
-	}
-
-	# read the pod a paragraph at a time
-	warn "Scanning for sections in input file(s)\n" if $Verbose;
-	$/ = "";
-	my @poddata = <POD>;
-	close(POD);
-
-	# be eol agnostic
-	for (@poddata) {
-		if (/\r/) {
-			if (/\r\n/) {
-				@poddata = map {
-					s/\r\n/\n/g;
-					/\n\n/
-						? map { "$_\n\n" } split /\n\n/
-						: $_
-				} @poddata;
-			} else {
-				@poddata = map {
-					s/\r/\n/g;
-					/\n\n/
-						? map { "$_\n\n" } split /\n\n/
-						: $_
-				} @poddata;
-			}
-			last;
-		}
-	}
-
-	clean_data( \@poddata );
-
-	# scan the pod for =head[1-6] directives and build an index
-	my $index = scan_headings( \%Sections, @poddata );
-
-	unless ($index) {
-		warn "No headings in $Podfile\n" if $Verbose;
-	}
-
-	# open the output file
-	open( HTML, ">$Htmlfile" )
-		|| die "$0: cannot open $Htmlfile file for output: $!\n";
-
-	# put a title in the HTML file if one wasn't specified
-	if ( $Title eq '' ) {
-	TITLE_SEARCH: {
-			for ( my $i = 0 ; $i < @poddata ; $i++ ) {
-				if ( $poddata[$i] =~ /^=head1\s*NAME\b/m ) {
-					for my $para ( @poddata[ $i, $i + 1 ] ) {
-						last TITLE_SEARCH
-							if ($Title) = $para =~ /(\S+\s+-+.*\S)/s;
-					}
-				}
-
-			}
-		}
-	}
-	if ( !$Title and $Podfile =~ /\.pod\z/ ) {
-
-		# probably a split pod so take first =head[12] as title
-		for ( my $i = 0 ; $i < @poddata ; $i++ ) {
-			last if ($Title) = $poddata[$i] =~ /^=head[12]\s*(.*)/;
-		}
-		warn "adopted '$Title' as title for $Podfile\n"
-			if $Verbose and $Title;
-	}
-	if ($Title) {
-		$Title =~ s/\s*\(.*\)//;
-	} else {
-		warn "$0: no title for $Podfile.\n" unless $Quiet;
-		$Podfile =~ /^(.*)(\.[^.\/]+)?\z/s;
-		$Title = ( $Podfile eq "-" ? 'No Title' : $1 );
-		warn "using $Title" if $Verbose;
-	}
-	$Title = html_escape($Title);
-
-	my $csslink   = '';
-	my $bodystyle = ' style="background-color: white"';
-	my $tdstyle   = ' style="background-color: #cccccc"';
-
-	if ($Css) {
-		$csslink = qq(\n<link rel="stylesheet" href="$Css" type="text/css" />);
-		$csslink =~ s,\\,/,g;
-		$csslink =~ s,(/.):,$1|,;
-		$bodystyle = '';
-		$tdstyle   = '';
-	}
-
-	my $block = $Header ? <<END_OF_BLOCK : '';
-<table border="0" width="100%" cellspacing="0" cellpadding="3">
-<tr><td class="block"$tdstyle valign="middle">
-<big><strong><span class="block">&nbsp;$Title</span></strong></big>
-</td></tr>
-</table>
-END_OF_BLOCK
-
-	print HTML <<END_OF_HEAD;
-<?xml version="1.0" ?>
-<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
-<html xmlns="http://www.w3.org/1999/xhtml">
-<head>
-<title>$Title</title>$csslink
-<meta http-equiv="content-type" content="text/html; charset=utf-8" />
-<link rev="made" href="mailto:$Config{perladmin}" />
-<style type="text/css">
-	.branch {
-		list-style: none;
-	}
-
-	li code {
-		padding-right: 0.5em;
-	}
-
-	.example {
-		width: 98%;
-		padding: 0.5em;
-		float: left;
-		font-family: monospace;
-	}
-
-	.example .line_number {
-		float: left;
-		text-align: right;
-		margin-right: 10px;
-		padding-right: 5px;
-		border-right: 1px solid white;
-	}
-
-	.example .content {
-		float: left;
-	}
-
-	.example .content pre {
-		margin: 0;
-	}
-
-	td > table {
-		margin: 0.5em;
-		border-collapse: collapse;
-	}
-
-	td > table td {
-		border: 1px solid black;
-	}
-
-	.synComment {
-		color: rgb(135,206,235);
-	}
-	.synPreProc {
-		color: rgb(205,92,92);
-	}
-	.synError {
-	}
-	.synConstant {
-		color: #ffa0a0;
-	}
-	.synSpecial {
-		color: rgb(255,222,173);
-	}
-	.synIgnore {
-		color: rgb(102,102,102);
-	}
-	.synNormal {
-		color: rgb(255,255,255);
-		background-color: rgb(51,51,51);
-	}
-	.synType {
-		color: rgb(189,183,107);
-	}
-	.synIdentifier {
-		color: rgb(152,251,152);
-	}
-	.synTodo {
-		color: rgb(255,69,0);
-		background-color: rgb(238,238,0);
-	}
-	.synStatement {
-		color: rgb(240,230,140);
-	}
-
-</style>
-</head>
-
-<body$bodystyle>
-$block
-END_OF_HEAD
-
-	# load/reload/validate/cache %Pages and %Items
-	get_cache( $Dircache, $Itemcache, \@Podpath, $Podroot, $Recurse );
-
-	# scan the pod for =item directives
-	scan_items( \%Local_Items, "", @poddata );
-
-	# put an index at the top of the file.  note, if $Doindex is 0 we
-	# still generate an index, but surround it with an html comment.
-	# that way some other program can extract it if desired.
-	$index =~ s/--+/-/g;
-
-	my $hr = ( $Doindex and $index ) ? qq(<hr />) : "";
-
-	unless ($Doindex) {
-		$index = qq(<!--\n$index\n-->\n);
-	}
-
-	print HTML << "END_OF_INDEX";
-
-<!-- INDEX BEGIN -->
-<div>
-<p><a name=\"__index__\"></a></p>
-$index
-$hr
-</div>
-<!-- INDEX END -->
-
-END_OF_INDEX
-
-	# now convert this file
-	my $after_item;    # set to true after an =item
-	my $need_dd = 0;
-	warn "Converting input file $Podfile\n" if $Verbose;
-	foreach my $i ( 0 .. $#poddata ) {
-		$_         = $poddata[$i];
-		$Paragraph = $i + 1;
-		if (/^(=.*)/s) {    # is it a pod directive?
-			$Ignore     = 0;
-			$after_item = 0;
-			$need_dd    = 0;
-			$_          = $1;
-			if (/^=begin\s+(\S+)\s*(.*)/si) {    # =begin
-				process_begin( $1, $2 );
-			} elsif (/^=end\s+(\S+)\s*(.*)/si) {    # =end
-				process_end( $1, $2 );
-			} elsif (/^=cut/) {                     # =cut
-				process_cut();
-			} elsif (/^=pod/) {                     # =pod
-				process_pod();
-			} else {
-				next if @Begin_Stack && $Begin_Stack[-1] ne 'html';
-
-				if (/^=(head[1-6])\s+(.*\S)/s) {    # =head[1-6] heading
-					process_head( $1, $2, $Doindex && $index );
-				} elsif (/^=item\s*(.*\S)?/sm) {    # =item text
-					$need_dd    = process_item($1);
-					$after_item = 1;
-				} elsif (/^=over\s*(.*)/) {         # =over N
-					process_over();
-				} elsif (/^=back/) {                # =back
-					process_back($need_dd);
-				} elsif (/^=for\s+(\S+)\s*(.*)/si) {    # =for
-					process_for( $1, $2 );
-				} else {
-					/^=(\S*)\s*/;
-					warn "$0: $Podfile: unknown pod directive '$1' in "
-						. "paragraph $Paragraph.  ignoring.\n"
-						unless $Quiet;
-				}
-			}
-			$Top = 0;
-		} else {
-			next if $Ignore;
-			next if @Begin_Stack && $Begin_Stack[-1] ne 'html';
-			print HTML and next if @Begin_Stack && $Begin_Stack[-1] eq 'html';
-			print HTML "<dd>\n" if $need_dd;
-			my $text = $_;
-			if ( $text =~ /\A\s+/ ) {
-				process_pre( \$text );
-				print HTML "<pre>\n$text</pre>\n";
-
-			} else {
-				process_text( \$text );
-
-				# experimental: check for a paragraph where all lines
-				# have some ...\t...\t...\n pattern
-				if ( $text =~ /\t/ ) {
-					my @lines = split( "\n", $text );
-					if ( @lines > 1 ) {
-						my $all = 2;
-						foreach my $line (@lines) {
-							if ( $line =~ /\S/ && $line !~ /\t/ ) {
-								$all--;
-								last if $all == 0;
-							}
-						}
-						if ( $all > 0 ) {
-							$text =~ s/\t+/<td>/g;
-							$text =~ s/^/<tr><td>/gm;
-							$text =
-								  '<table cellspacing="0" cellpadding="0">'
-								. $text
-								. '</table>';
-						}
-					}
-				}
-				## end of experimental
-
-				if ($after_item) {
-					$After_Lpar = 1;
-				}
-				print HTML "<p>$text</p>\n";
-			}
-			print HTML "</dd>\n" if $need_dd;
-			$after_item = 0;
-		}
-	}
-
-	# finish off any pending directives
-	finish_list();
-
-	# link to page index
-	print HTML "<p><a href=\"#__index__\"><small>$Backlink</small></a></p>\n"
-		if $Doindex
-			and $index
-			and $Backlink;
-
-	print HTML <<END_OF_TAIL;
-$block
-</body>
-
-</html>
-END_OF_TAIL
-
-	# close the html file
-	close(HTML);
-
-	warn "Finished\n" if $Verbose;
-}
-
-##############################################################################
-
-sub usage {
-	my $podfile = shift;
-	warn "$0: $podfile: @_\n" if @_;
-	die <<END_OF_USAGE;
-Usage:  $0 --help --htmlroot=<name> --infile=<name> --outfile=<name>
-           --podpath=<name>:...:<name> --podroot=<name>
-           --libpods=<name>:...:<name> --recurse --verbose --index
-           --netscape --norecurse --noindex --cachedir=<name>
-
-  --backlink     - set text for "back to top" links (default: none).
-  --cachedir     - directory for the item and directory cache files.
-  --css          - stylesheet URL
-  --flush        - flushes the item and directory caches.
-  --[no]header   - produce block header/footer (default is no headers).
-  --help         - prints this message.
-  --hiddendirs   - search hidden directories in podpath
-  --htmldir      - directory for resulting HTML files.
-  --htmlroot     - http-server base directory from which all relative paths
-                   in podpath stem (default is /).
-  --[no]index    - generate an index at the top of the resulting html
-                   (default behaviour).
-  --infile       - filename for the pod to convert (input taken from stdin
-                   by default).
-  --libpods      - colon-separated list of pages to search for =item pod
-                   directives in as targets of C<> and implicit links (empty
-                   by default).  note, these are not filenames, but rather
-                   page names like those that appear in L<> links.
-  --outfile      - filename for the resulting html file (output sent to
-                   stdout by default).
-  --podpath      - colon-separated list of directories containing library
-                   pods (empty by default).
-  --podroot      - filesystem base directory from which all relative paths
-                   in podpath stem (default is .).
-  --[no]quiet    - suppress some benign warning messages (default is off).
-  --[no]recurse  - recurse on those subdirectories listed in podpath
-                   (default behaviour).
-  --title        - title that will appear in resulting html file.
-  --[no]verbose  - self-explanatory (off by default).
-  --[no]netscape - deprecated, has no effect. for backwards compatibility only.
-
-END_OF_USAGE
-
-}
-
-sub parse_command_line {
-	my (
-		$opt_backlink, $opt_cachedir, $opt_css,     $opt_flush,
-		$opt_header,   $opt_help,     $opt_htmldir, $opt_htmlroot,
-		$opt_index,    $opt_infile,   $opt_libpods, $opt_netscape,
-		$opt_outfile,  $opt_podpath,  $opt_podroot, $opt_quiet,
-		$opt_recurse,  $opt_title,    $opt_verbose, $opt_hiddendirs
-	);
-
-	unshift @ARGV, split ' ', $Config{pod2html} if $Config{pod2html};
-	my $result = GetOptions(
-		'backlink=s'  => \$opt_backlink,
-		'cachedir=s'  => \$opt_cachedir,
-		'css=s'       => \$opt_css,
-		'flush'       => \$opt_flush,
-		'header!'     => \$opt_header,
-		'help'        => \$opt_help,
-		'hiddendirs!' => \$opt_hiddendirs,
-		'htmldir=s'   => \$opt_htmldir,
-		'htmlroot=s'  => \$opt_htmlroot,
-		'index!'      => \$opt_index,
-		'infile=s'    => \$opt_infile,
-		'libpods=s'   => \$opt_libpods,
-		'netscape!'   => \$opt_netscape,
-		'outfile=s'   => \$opt_outfile,
-		'podpath=s'   => \$opt_podpath,
-		'podroot=s'   => \$opt_podroot,
-		'quiet!'      => \$opt_quiet,
-		'recurse!'    => \$opt_recurse,
-		'title=s'     => \$opt_title,
-		'verbose!'    => \$opt_verbose,
-	);
-	usage( "-", "invalid parameters" ) if not $result;
-
-	usage("-") if defined $opt_help;    # see if the user asked for help
-	$opt_help = "";                     # just to make -w shut-up.
-
-	@Podpath = split( ":", $opt_podpath ) if defined $opt_podpath;
-	@Libpods = split( ":", $opt_libpods ) if defined $opt_libpods;
-
-	$Backlink   = $opt_backlink   if defined $opt_backlink;
-	$Cachedir   = $opt_cachedir   if defined $opt_cachedir;
-	$Css        = $opt_css        if defined $opt_css;
-	$Header     = $opt_header     if defined $opt_header;
-	$Htmldir    = $opt_htmldir    if defined $opt_htmldir;
-	$Htmlroot   = $opt_htmlroot   if defined $opt_htmlroot;
-	$Doindex    = $opt_index      if defined $opt_index;
-	$Podfile    = $opt_infile     if defined $opt_infile;
-	$HiddenDirs = $opt_hiddendirs if defined $opt_hiddendirs;
-	$Htmlfile   = $opt_outfile    if defined $opt_outfile;
-	$Podroot    = $opt_podroot    if defined $opt_podroot;
-	$Quiet      = $opt_quiet      if defined $opt_quiet;
-	$Recurse    = $opt_recurse    if defined $opt_recurse;
-	$Title      = $opt_title      if defined $opt_title;
-	$Verbose    = $opt_verbose    if defined $opt_verbose;
-
-	warn "Flushing item and directory caches\n"
-		if $opt_verbose && defined $opt_flush;
-	$Dircache  = "$Cachedir/pod2htmd.tmp";
-	$Itemcache = "$Cachedir/pod2htmi.tmp";
-	if ( defined $opt_flush ) {
-		1 while unlink( $Dircache, $Itemcache );
-	}
-}
-
-my $Saved_Cache_Key;
-
-sub get_cache {
-	my ( $dircache, $itemcache, $podpath, $podroot, $recurse ) = @_;
-	my @cache_key_args = @_;
-
-	# A first-level cache:
-	# Don't bother reading the cache files if they still apply
-	# and haven't changed since we last read them.
-
-	my $this_cache_key = cache_key(@cache_key_args);
-
-	return if $Saved_Cache_Key and $this_cache_key eq $Saved_Cache_Key;
-
-	# load the cache of %Pages and %Items if possible.  $tests will be
-	# non-zero if successful.
-	my $tests = 0;
-	if ( -f $dircache && -f $itemcache ) {
-		warn "scanning for item cache\n" if $Verbose;
-		$tests = load_cache( $dircache, $itemcache, $podpath, $podroot );
-	}
-
-	# if we didn't succeed in loading the cache then we must (re)build
-	#  %Pages and %Items.
-	if ( !$tests ) {
-		warn "scanning directories in pod-path\n" if $Verbose;
-		scan_podpath( $podroot, $recurse, 0 );
-	}
-	$Saved_Cache_Key = cache_key(@cache_key_args);
-}
-
-sub cache_key {
-	my ( $dircache, $itemcache, $podpath, $podroot, $recurse ) = @_;
-	return join( '!',
-		$dircache, $itemcache, $recurse, @$podpath, $podroot, stat($dircache),
-		stat($itemcache) );
-}
-
-#
-# load_cache - tries to find if the caches stored in $dircache and $itemcache
-#  are valid caches of %Pages and %Items.  if they are valid then it loads
-#  them and returns a non-zero value.
-#
-sub load_cache {
-	my ( $dircache, $itemcache, $podpath, $podroot ) = @_;
-	my ($tests);
-	local $_;
-
-	$tests = 0;
-
-	open( CACHE, "<$itemcache" )
-		|| die "$0: error opening $itemcache for reading: $!\n";
-	$/ = "\n";
-
-	# is it the same podpath?
-	$_ = <CACHE>;
-	chomp($_);
-	$tests++ if ( join( ":", @$podpath ) eq $_ );
-
-	# is it the same podroot?
-	$_ = <CACHE>;
-	chomp($_);
-	$tests++ if ( $podroot eq $_ );
-
-	# load the cache if its good
-	if ( $tests != 2 ) {
-		close(CACHE);
-		return 0;
-	}
-
-	warn "loading item cache\n" if $Verbose;
-	while (<CACHE>) {
-		/(.*?) (.*)$/;
-		$Items{$1} = $2;
-	}
-	close(CACHE);
-
-	warn "scanning for directory cache\n" if $Verbose;
-	open( CACHE, "<$dircache" )
-		|| die "$0: error opening $dircache for reading: $!\n";
-	$/     = "\n";
-	$tests = 0;
-
-	# is it the same podpath?
-	$_ = <CACHE>;
-	chomp($_);
-	$tests++ if ( join( ":", @$podpath ) eq $_ );
-
-	# is it the same podroot?
-	$_ = <CACHE>;
-	chomp($_);
-	$tests++ if ( $podroot eq $_ );
-
-	# load the cache if its good
-	if ( $tests != 2 ) {
-		close(CACHE);
-		return 0;
-	}
-
-	warn "loading directory cache\n" if $Verbose;
-	while (<CACHE>) {
-		/(.*?) (.*)$/;
-		$Pages{$1} = $2;
-	}
-
-	close(CACHE);
-
-	return 1;
-}
-
-#
-# scan_podpath - scans the directories specified in @podpath for directories,
-#  .pod files, and .pm files.  it also scans the pod files specified in
-#  @Libpods for =item directives.
-#
-sub scan_podpath {
-	my ( $podroot, $recurse, $append ) = @_;
-	my ( $pwd, $dir );
-	my ( $libpod, $dirname, $pod, @files, @poddata );
-
-	unless ($append) {
-		%Items = ();
-		%Pages = ();
-	}
-
-	# scan each directory listed in @Podpath
-	$pwd = getcwd();
-	chdir($podroot)
-		|| die "$0: error changing to directory $podroot: $!\n";
-	foreach $dir (@Podpath) {
-		scan_dir( $dir, $recurse );
-	}
-
-	# scan the pods listed in @Libpods for =item directives
-	foreach $libpod (@Libpods) {
-
-		# if the page isn't defined then we won't know where to find it
-		# on the system.
-		next unless defined $Pages{$libpod} && $Pages{$libpod};
-
-		# if there is a directory then use the .pod and .pm files within it.
-		# NOTE: Only finds the first so-named directory in the tree.
-		#	if ($Pages{$libpod} =~ /([^:]*[^(\.pod|\.pm)]):/) {
-		if ( $Pages{$libpod} =~ /([^:]*(?<!\.pod)(?<!\.pm)):/ ) {
-
-			#  find all the .pod and .pm files within the directory
-			$dirname = $1;
-			opendir( DIR, $dirname )
-				|| die "$0: error opening directory $dirname: $!\n";
-			@files = grep( /(\.pod|\.pm)\z/ && !-d $_, readdir(DIR) );
-			closedir(DIR);
-
-			# scan each .pod and .pm file for =item directives
-			foreach $pod (@files) {
-				open( POD, "<$dirname/$pod" )
-					|| die "$0: error opening $dirname/$pod for input: $!\n";
-				@poddata = <POD>;
-				close(POD);
-				clean_data( \@poddata );
-
-				scan_items( \%Items, "$dirname/$pod", @poddata );
-			}
-
-			# use the names of files as =item directives too.
-### Don't think this should be done this way - confuses issues.(WL)
-###	    foreach $pod (@files) {
-###		$pod =~ /^(.*)(\.pod|\.pm)$/;
-###		$Items{$1} = "$dirname/$1.html" if $1;
-###	    }
-		} elsif ( $Pages{$libpod} =~ /([^:]*\.pod):/
-			|| $Pages{$libpod} =~ /([^:]*\.pm):/ )
-		{
-
-			# scan the .pod or .pm file for =item directives
-			$pod = $1;
-			open( POD, "<$pod" )
-				|| die "$0: error opening $pod for input: $!\n";
-			@poddata = <POD>;
-			close(POD);
-			clean_data( \@poddata );
-
-			scan_items( \%Items, "$pod", @poddata );
-		} else {
-			warn "$0: shouldn't be here (line " . __LINE__ . "\n" unless $Quiet;
-		}
-	}
-	@poddata = ();    # clean-up a bit
-
-	chdir($pwd)
-		|| die "$0: error changing to directory $pwd: $!\n";
-
-	# cache the item list for later use
-	warn "caching items for later use\n" if $Verbose;
-	open( CACHE, ">$Itemcache" )
-		|| die "$0: error open $Itemcache for writing: $!\n";
-
-	print CACHE join( ":", @Podpath ) . "\n$podroot\n";
-	foreach my $key ( keys %Items ) {
-		print CACHE "$key $Items{$key}\n";
-	}
-
-	close(CACHE);
-
-	# cache the directory list for later use
-	warn "caching directories for later use\n" if $Verbose;
-	open( CACHE, ">$Dircache" )
-		|| die "$0: error open $Dircache for writing: $!\n";
-
-	print CACHE join( ":", @Podpath ) . "\n$podroot\n";
-	foreach my $key ( keys %Pages ) {
-		print CACHE "$key $Pages{$key}\n";
-	}
-
-	close(CACHE);
-}
-
-#
-# scan_dir - scans the directory specified in $dir for subdirectories, .pod
-#  files, and .pm files.  notes those that it finds.  this information will
-#  be used later in order to figure out where the pages specified in L<>
-#  links are on the filesystem.
-#
-sub scan_dir {
-	my ( $dir, $recurse ) = @_;
-	my ( $t, @subdirs, @pods, $pod, $dirname, @dirs );
-	local $_;
-
-	@subdirs = ();
-	@pods    = ();
-
-	opendir( DIR, $dir )
-		|| die "$0: error opening directory $dir: $!\n";
-	while ( defined( $_ = readdir(DIR) ) ) {
-		if (   -d "$dir/$_"
-			&& $_ ne "."
-			&& $_ ne ".."
-			&& ( $HiddenDirs || !/^\./ ) )
-		{    # directory
-			$Pages{$_} = "" unless defined $Pages{$_};
-			$Pages{$_} .= "$dir/$_:";
-			push( @subdirs, $_ );
-		} elsif (/\.pod\z/) {    # .pod
-			s/\.pod\z//;
-			$Pages{$_} = "" unless defined $Pages{$_};
-			$Pages{$_} .= "$dir/$_.pod:";
-			push( @pods, "$dir/$_.pod" );
-		} elsif (/\.html\z/) {    # .html
-			s/\.html\z//;
-			$Pages{$_} = "" unless defined $Pages{$_};
-			$Pages{$_} .= "$dir/$_.pod:";
-		} elsif (/\.pm\z/) {      # .pm
-			s/\.pm\z//;
-			$Pages{$_} = "" unless defined $Pages{$_};
-			$Pages{$_} .= "$dir/$_.pm:";
-			push( @pods, "$dir/$_.pm" );
-		} elsif ( -T "$dir/$_" ) {    # script(?)
-			local *F;
-			if ( open( F, "$dir/$_" ) ) {
-				my $line;
-				while ( defined( $line = <F> ) ) {
-					if ( $line =~ /^=(?:pod|head1)/ ) {
-						$Pages{$_} = "" unless defined $Pages{$_};
-						$Pages{$_} .= "$dir/$_.pod:";
-						last;
-					}
-				}
-				close(F);
-			}
-		}
-	}
-	closedir(DIR);
-
-	# recurse on the subdirectories if necessary
-	if ($recurse) {
-		foreach my $subdir (@subdirs) {
-			scan_dir( "$dir/$subdir", $recurse );
-		}
-	}
-}
-
-#
-# scan_headings - scan a pod file for head[1-6] tags, note the tags, and
-#  build an index.
-#
-sub scan_headings {
-	my ( $sections, @data ) = @_;
-	my ( $tag, $which_head, $otitle, $listdepth, $index );
-
-	local $Ignore = 0;
-
-	$listdepth = 0;
-	$index     = "";
-
-	# scan for =head directives, note their name, and build an index
-	#  pointing to each of them.
-	foreach my $line (@data) {
-		if ( $line =~ /^=(head)([1-6])\s+(.*)/ ) {
-			( $tag, $which_head, $otitle ) = ( $1, $2, $3 );
-
-			my $title = depod($otitle);
-			my $name  = anchorify($title);
-			$$sections{$name} = 1;
-			$title = process_text( \$otitle );
-
-			while ( $which_head != $listdepth ) {
-				if ( $which_head > $listdepth ) {
-					$index .= "\n"
-						. ( "\t" x ($listdepth) )
-						. ( $listdepth > 0 ? qq{<li class="branch">\n}
-							. "\t"x($listdepth + 1): "" )
-						. "<ul>";
-					$listdepth++;
-				} elsif ( $which_head < $listdepth ) {
-					$listdepth--;
-					$index .= "\n"
-						. ( "\t" x $listdepth ) 
-						. ( $listdepth > 0 ? "\t" : "" )
-						. "</ul>"
-						. ( $listdepth >= 0 ? "\n" . ("\t"x$listdepth)
-							. "</li>" : "" )
-						. "\n";
-				}
-			}
-
-			$index .= "\n"
-				. ( "\t" x $listdepth ) . "<li>"
-				. "<a href=\"#"
-				. $name . "\">"
-				. $title
-				. "</a></li>";
-		}
-	}
-
-	# finish off the lists
-	while ( $listdepth-- ) {
-		$index .= "\n" . ( "\t" x $listdepth )
-			. ($listdepth > 0 ? "\t" : "")
-			."</ul>\n"
-			. ($listdepth > 0 ? ("\t" x $listdepth) . "</li>" : "" );
-	}
-
-	# get rid of bogus lists
-	$index =~ s,\t*<ul>\s*</ul>\n,,g;
-
-	return $index;
-}
-
-#
-# scan_items - scans the pod specified by $pod for =item directives.  we
-#  will use this information later on in resolving C<> links.
-#
-sub scan_items {
-	my ( $itemref, $pod, @poddata ) = @_;
-	my ( $i, $item );
-	local $_;
-
-	$pod =~ s/\.pod\z//;
-	$pod .= ".html" if $pod;
-
-	foreach $i ( 0 .. $#poddata ) {
-		my $txt = depod( $poddata[$i] );
-
-		# figure out what kind of item it is.
-		# Build string for referencing this item.
-		if ( $txt =~ /\A=item\s+\*\s*(.*)\Z/s ) {    # bullet
-			next unless $1;
-			$item = $1;
-		} elsif ( $txt =~ /\A=item\s+(?>\d+\.?)\s*(.*)\Z/s ) {   # numbered list
-			$item = $1;
-		} elsif ( $txt =~ /\A=item\s+(.*)\Z/s ) {                # plain item
-			$item = $1;
-		} else {
-			next;
-		}
-		my $fid = fragment_id($item);
-		$$itemref{$fid} = "$pod" if $fid;
-	}
-}
-
-#
-# process_head - convert a pod head[1-6] tag and convert it to HTML format.
-#
-sub process_head {
-	my ( $tag, $heading, $hasindex ) = @_;
-
-	# figure out the level of the =head
-	$tag =~ /head([1-6])/;
-	my $level = $1;
-
-	if ($Listlevel) {
-		warn
-"$0: $Podfile: unterminated list at =head in paragraph $Paragraph.  ignoring.\n"
-			unless $Quiet;
-		while ($Listlevel) {
-			process_back();
-		}
-	}
-
-	print HTML "<p>\n";
-	if ( $level == 1 && !$Top ) {
-		print HTML "<a href=\"#__index__\"><small>$Backlink</small></a>\n"
-			if $hasindex and $Backlink;
-		print HTML "</p>\n<hr />\n";
-	} else {
-		print HTML "</p>\n";
-	}
-
-	my $name    = anchorify( depod($heading) );
-	my $convert = process_text( \$heading );
-	$convert =~ s{</?a[^>]+>}{}g;
-	print HTML "<h$level><a name=\"$name\" />$convert</h$level>\n";
-}
-
-#
-# emit_item_tag - print an =item's text
-# Note: The global $EmittedItem is used for inhibiting self-references.
-#
-my $EmittedItem;
-
-sub emit_item_tag($$$) {
-	my ( $otext, $text, $compact ) = @_;
-	my $item = fragment_id( depod($text), -generate );
-	Carp::confess( "Undefined fragment '$text' ("
-			. depod($text)
-			. ") from fragment_id() in emit_item_tag() in $Podfile" )
-		if !defined $item;
-	$EmittedItem = $item;
-	### print STDERR "emit_item_tag=$item ($text)\n";
-
-	print HTML '<strong>';
-	if ( $Items_Named{$item}++ ) {
-		print HTML process_text( \$otext );
-	} else {
-		my $name = $item;
-		$name = anchorify($name);
-		print HTML 
-#qq{<a name="$name" class="item">},
-		process_text( \$otext ),
-		#	'</a>'
-		;
-	}
-	print HTML "</strong>\n";
-	undef($EmittedItem);
-}
-
-sub emit_li {
-	my ($tag) = @_;
-	if ( $Items_Seen[$Listlevel]++ == 0 ) {
-		push( @Listend, "</$tag>" );
-		print HTML "<$tag>\n";
-	}
-	my $emitted = $tag eq 'dl' ? 'dt' : 'li';
-	print HTML "<$emitted>";
-	return $emitted;
-}
-
-#
-# process_item - convert a pod item tag and convert it to HTML format.
-#
-sub process_item {
-	my ($otext) = @_;
-	my $need_dd = 0;    # set to 1 if we need a <dd></dd> after an item
-
-	# lots of documents start a list without doing an =over.  this is
-	# bad!  but, the proper thing to do seems to be to just assume
-	# they did do an =over.  so warn them once and then continue.
-	if ( $Listlevel == 0 ) {
-		warn
-"$0: $Podfile: unexpected =item directive in paragraph $Paragraph.  ignoring.\n"
-			unless $Quiet;
-		process_over();
-	}
-
-	# formatting: insert a paragraph if preceding item has >1 paragraph
-	if ($After_Lpar) {
-		print HTML $need_dd ? "</dd>\n" : "</li>\n" if $After_Lpar;
-		$After_Lpar = 0;
-	}
-
-	# remove formatting instructions from the text
-	my $text = depod($otext);
-
-	my $emitted;    # the tag actually emitted, used for closing
-
-	# all the list variants:
-	if ( $text =~ /\A\*/ ) {    # bullet
-		$emitted = emit_li('ul');
-		if ( $text =~ /\A\*\s+(\S.*)\Z/s ) {    # with additional text
-			my $tag = $1;
-			$otext =~ s/\A\*\s+//;
-			emit_item_tag( $otext, $tag, 1 );
-		}
-		print HTML "</li>"
-	} elsif ( $text =~ /\A\d+/ ) {    # numbered list
-		$emitted = emit_li('ol');
-		if ( $text =~ /\A(?>\d+\.?)\s*(\S.*)\Z/s ) {    # with additional text
-			my $tag = $1;
-			$otext =~ s/\A\d+\.?\s*//;
-			emit_item_tag( $otext, $tag, 1 );
-		}
-		print HTML "</li>";
-	} else {    # definition list
-		$emitted = emit_li('dl');
-		if ( $text =~ /\A(.+)\Z/s ) {    # should have text
-			emit_item_tag( $otext, $text, 1 );
-		}
-		$need_dd = 1;
-	}
-	print HTML "\n";
-	return $need_dd;
-}
-
-#
-# process_over - process a pod over tag and start a corresponding HTML list.
-#
-sub process_over {
-
-	# start a new list
-	$Listlevel++;
-	push( @Items_Seen, 0 );
-	$After_Lpar = 0;
-}
-
-#
-# process_back - process a pod back tag and convert it to HTML format.
-#
-sub process_back {
-	my $need_dd = shift;
-	if ( $Listlevel == 0 ) {
-		warn
-"$0: $Podfile: unexpected =back directive in paragraph $Paragraph.  ignoring.\n"
-			unless $Quiet;
-		return;
-	}
-
-	# close off the list.  note, I check to see if $Listend[$Listlevel] is
-	# defined because an =item directive may have never appeared and thus
-	# $Listend[$Listlevel] may have never been initialized.
-	$Listlevel--;
-	if ( defined $Listend[$Listlevel] ) {
-		print HTML $need_dd ? "</dd>\n" : "</li>\n" if $After_Lpar;
-		print HTML $Listend[$Listlevel];
-		print HTML "\n";
-		pop(@Listend);
-	}
-	$After_Lpar = 0;
-
-	# clean up item count
-	pop(@Items_Seen);
-}
-
-#
-# process_cut - process a pod cut tag, thus start ignoring pod directives.
-#
-sub process_cut {
-	$Ignore = 1;
-}
-
-#
-# process_pod - process a pod tag, thus stop ignoring pod directives
-# until we see a corresponding cut.
-#
-sub process_pod {
-
-	# no need to set $Ignore to 0 cause the main loop did it
-}
-
-#
-# process_for - process a =for pod tag.  if it's for html, spit
-# it out verbatim, if illustration, center it, otherwise ignore it.
-#
-sub process_for {
-	my ( $whom, $text ) = @_;
-	if ( $whom =~ /^(pod2)?html$/i ) {
-		print HTML $text;
-	} elsif ( $whom =~ /^illustration$/i ) {
-		1 while chomp $text;
-		for my $ext (qw[.png .gif .jpeg .jpg .tga .pcl .bmp]) {
-			$text .= $ext, last if -r "$text$ext";
-		}
-		print HTML
-qq{<p align="center"><img src="$text" alt="$text illustration" /></p>};
-	}
-}
-
-#
-# process_begin - process a =begin pod tag.  this pushes
-# whom we're beginning on the begin stack.  if there's a
-# begin stack, we only print if it us.
-#
-sub process_begin {
-	my ( $whom, $text ) = @_;
-	$whom = lc($whom);
-	push( @Begin_Stack, $whom );
-	if ( $whom =~ /^(pod2)?html$/ ) {
-		print HTML $text if $text;
-	}
-}
-
-#
-# process_end - process a =end pod tag.  pop the
-# begin stack.  die if we're mismatched.
-#
-sub process_end {
-	my ( $whom, $text ) = @_;
-	$whom = lc($whom);
-	if ( !defined $Begin_Stack[-1] or $Begin_Stack[-1] ne $whom ) {
-		Carp::confess(
-			"Unmatched begin/end at chunk $Paragraph in pod $Podfile\n");
-	}
-	pop(@Begin_Stack);
-}
-
-#
-# process_pre - indented paragraph, made into <pre></pre>
-#
-sub process_pre {
-	my ($text) = @_;
-	my ($rest);
-	return if $Ignore;
-
-	$rest = $$text;
-
-	# insert spaces in place of tabs
-	$rest =~ s#(.+)#
-	    my $line = $1;
-            1 while $line =~ s/(\t+)/' ' x ((length($1) * 8) - $-[0] % 8)/e;
-	    $line;
-	#eg;
-
-	# convert some special chars to HTML escapes
-	$rest = html_escape($rest);
-
-	# try and create links for all occurrences of perl.* within
-	# the preformatted text.
-	$rest =~ s{
-	         (\s*)(perl\w+)
-	      }{
-		 if ( defined $Pages{$2} ){	# is a link
-		     qq($1<a href="$Htmlroot/$Pages{$2}">$2</a>);
-		 } elsif (defined $Pages{dosify($2)}) {	# is a link
-		     qq($1<a href="$Htmlroot/$Pages{dosify($2)}">$2</a>);
-		 } else {
-		     "$1$2";
-		 }
-	      }xeg;
-	$rest =~ s{
-		 (<a\ href="?) ([^>:]*:)? ([^>:]*) \.pod: ([^>:]*:)?
-               }{
-                  my $url ;
-                  if ( $Htmlfileurl ne '' ){
-		     # Here, we take advantage of the knowledge
-		     # that $Htmlfileurl ne '' implies $Htmlroot eq ''.
-		     # Since $Htmlroot eq '', we need to prepend $Htmldir
-		     # on the fron of the link to get the absolute path
-		     # of the link's target. We check for a leading '/'
-		     # to avoid corrupting links that are #, file:, etc.
-		     my $old_url = $3 ;
-		     $old_url = "$Htmldir$old_url" if $old_url =~ m{^\/};
- 		     $url = relativize_url( "$old_url.html", $Htmlfileurl );
-	          } else {
-		     $url = "$3.html" ;
-		  }
-		  "$1$url" ;
-	       }xeg;
-
-	# Look for embedded URLs and make them into links.  We don't
-	# relativize them since they are best left as the author intended.
-
-	my $urls = '(' . join(
-		'|', qw{
-			http
-			telnet
-			mailto
-			news
-			gopher
-			file
-			wais
-			ftp
-			}
-	) . ')';
-
-	my $ltrs = '\w';
-	my $gunk = '/#~:.?+=&%@!\-';
-	my $punc = '.:!?\-;';
-	my $any  = "${ltrs}${gunk}${punc}";
-
-	$rest =~ s{
-	\b			# start at word boundary
-	(			# begin $1  {
-	    $urls :		# need resource and a colon
-	    (?!:)		# Ignore File::, among others.
-	    [$any] +?		# followed by one or more of any valid
-				#   character, but be conservative and
-				#   take only what you need to....
-	)			# end   $1  }
-	(?=
-	    &quot; &gt;		# maybe pre-quoted '<a href="...">'
-	|			# or:
-	    [$punc]*		# 0 or more punctuation
-	    (?:			#   followed
-		[^$any]		#   by a non-url char
-	    |			#   or
-		$		#   end of the string
-	    )			#
-	|			# or else
-	    $			#   then end of the string
-        )
-      }{<a href="$1">$1</a>}igox;
-
-	# text should be as it is (verbatim)
-	$$text = $rest;
-}
-
-#
-# pure text processing
-#
-# pure_text/inIS_text: differ with respect to automatic C<> recognition.
-# we don't want this to happen within IS
-#
-sub pure_text($) {
-	my $text = shift();
-	process_puretext( $text, 1 );
-}
-
-sub inIS_text($) {
-	my $text = shift();
-	process_puretext( $text, 0 );
-}
-
-#
-# process_puretext - process pure text (without pod-escapes) converting
-#  double-quotes and handling implicit C<> links.
-#
-sub process_puretext {
-	my ( $text, $notinIS ) = @_;
-
-	## Guessing at func() or [\$\@%&]*var references in plain text is destined
-	## to produce some strange looking ref's. uncomment to disable:
-	## $notinIS = 0;
-
-	my ( @words, $lead, $trail );
-
-	# keep track of leading and trailing white-space
-	$lead  = ( $text =~ s/\A(\s+)//s ? $1 : "" );
-	$trail = ( $text =~ s/(\s+)\Z//s ? $1 : "" );
-
-	# split at space/non-space boundaries
-	@words = split( /(?<=\s)(?=\S)|(?<=\S)(?=\s)/, $text );
-
-	# process each word individually
-	foreach my $word (@words) {
-
-		# skip space runs
-		next if $word =~ /^\s*$/;
-
-		# see if we can infer a link or a function call
-		#
-		# NOTE: This is a word based search, it won't automatically
-		# mark "substr($var, 1, 2)" because the 1st word would be "substr($var"
-		# User has to enclose those with proper C<>
-
-		if (
-			   $notinIS
-			&& $word =~ m/
-		^([a-z_]{2,})                 # The function name
-		\(
-		    ([0-9][a-z]*              # Manual page(1) or page(1M)
-		    |[^)]*[\$\@\%][^)]+       # ($foo), (1, @foo), (%hash)
-		    |                         # ()
-		    )
-		\)
-		([.,;]?)$                     # a possible punctuation follows
-	    /xi
-			)
-		{
-
-			# has parenthesis so should have been a C<> ref
-			## try for a pagename (perlXXX(1))?
-			my ( $func, $args, $rest ) = ( $1, $2, $3 || '' );
-			if ( $args =~ /^\d+$/ ) {
-				my $url = page_sect( $word, '' );
-				if ( defined $url ) {
-					$word =
-qq(<a href="$url" class="man">the $word manpage</a>$rest);
-					next;
-				}
-			}
-			## try function name for a link, append tt'ed argument list
-			$word = emit_C( $func, '', "($args)" ) . $rest;
-
-#### disabled. either all (including $\W, $\w+{.*} etc.) or nothing.
-##      } elsif( $notinIS && $word =~ /^[\$\@%&*]+\w+$/) {
-##	    # perl variables, should be a C<> ref
-##	    $word = emit_C( $word );
-
-		} elsif ( $word =~ m,^\w+://\w, ) {
-
-			# looks like a URL
-			# Don't relativize it: leave it as the author intended
-			$word = qq(<a href="$word">$word</a>);
-		} elsif ( $word =~ /[\w.-]+\@[\w-]+\.\w/ ) {
-
-			# looks like an e-mail address
-			my ( $w1, $w2, $w3 ) = ( "", $word, "" );
-			( $w1, $w2, $w3 ) = ( "(", $1, ")$2" ) if $word =~ /^\((.*?)\)(,?)/;
-			( $w1, $w2, $w3 ) = ( "&lt;", $1, "&gt;$2" )
-				if $word =~ /^<(.*?)>(,?)/;
-			$word = qq($w1<a href="mailto:$w2">$w2</a>$w3);
-		} else {
-			$word = html_escape($word) if $word =~ /["&<>]/;
-		}
-	}
-
-	# put everything back together
-	return $lead . join( '', @words ) . $trail;
-}
-
-#
-# process_text - handles plaintext that appears in the input pod file.
-# there may be pod commands embedded within the text so those must be
-# converted to html commands.
-#
-
-sub process_text1($$;$$);
-sub pattern ($) { $_[0] ? '\s+' . ( '>' x ( $_[0] + 1 ) ) : '>' }
-sub closing ($) { local ($_) = shift; ( defined && s/\s+\z// ) ? length : 0 }
-
-sub process_text {
-	return if $Ignore;
-	my ($tref) = @_;
-	my $res = process_text1( 0, $tref );
-	$res =~ s/\s+$//s;
-	$$tref = $res;
-}
-
-sub process_text_rfc_links {
-	my $text = shift;
-
-	# For every "RFCnnnn" or "RFC nnn", link it to the authoritative
-	# ource. Do not use the /i modifier here. Require "RFC" to be written in
-	#  in capital letters.
-
-	$text =~ s{
-	(?<=[^<>[:alpha:]])           # Make sure this is not an URL already
-	(RFC\s*([0-9]{1,5}))(?![0-9]) # max 5 digits
-    }
-    {<a href="http://www.ietf.org/rfc/rfc$2.txt" class="rfc">$1</a>}gx;
-
-	$text;
-}
-
-sub process_text1($$;$$) {
-	my ( $lev, $rstr, $func, $closing ) = @_;
-	my $res = '';
-
-	unless ( defined $func ) {
-		$func = '';
-		$lev++;
-	}
-
-	if ( $func eq 'B' ) {
-
-		# B<text> - boldface
-		$res = '<strong>' . process_text1( $lev, $rstr ) . '</strong>';
-
-	} elsif ( $func eq 'C' ) {
-
-		# C<code> - can be a ref or <code></code>
-		# need to extract text
-		my $par = go_ahead( $rstr, 'C', $closing );
-
-		## clean-up of the link target
-		my $text = depod($par);
-
-		### my $x = $par =~ /[BI]</ ? 'yes' : 'no' ;
-		### print STDERR "-->call emit_C($par) lev=$lev, par with BI=$x\n";
-
-		$res = emit_C( $text, $lev > 1 || ( $par =~ /[BI]</ ) );
-
-	} elsif ( $func eq 'E' ) {
-
-		# E<x> - convert to character
-		$$rstr =~ s/^([^>]*)>//;
-		my $escape = $1;
-		$escape =~ s/^(\d+|X[\dA-F]+)$/#$1/i;
-		$res = "&$escape;";
-
-	} elsif ( $func eq 'F' ) {
-
-		# F<filename> - italicize
-		$res = '<em class="file">' . process_text1( $lev, $rstr ) . '</em>';
-
-	} elsif ( $func eq 'I' ) {
-
-		# I<text> - italicize
-		$res = '<em>' . process_text1( $lev, $rstr ) . '</em>';
-
-	} elsif ( $func eq 'L' ) {
-
-		# L<link> - link
-		## L<text|cross-ref> => produce text, use cross-ref for linking
-		## L<cross-ref> => make text from cross-ref
-		## need to extract text
-		my $par = go_ahead( $rstr, 'L', $closing );
-
-		# some L<>'s that shouldn't be:
-		# a) full-blown URL's are emitted as-is
-		if ( $par =~ m{^\w+://}s ) {
-			return make_URL_href($par);
-		}
-
-		# b) C<...> is stripped and treated as C<>
-		if ( $par =~ /^C<(.*)>$/ ) {
-			my $text = depod($1);
-			return emit_C( $text, $lev > 1 || ( $par =~ /[BI]</ ) );
-		}
-
-		# analyze the contents
-		$par =~ s/\n/ /g;    # undo word-wrapped tags
-		my $opar = $par;
-		my $linktext;
-		if ( $par =~ s{^([^|]+)\|}{} ) {
-			$linktext = $1;
-		}
-
-		if( $par =~ m{^\w+://}s ) {
-			return make_URL_href( $par, $linktext );
-		}
-
-		# make sure sections start with a /
-		$par =~ s{^"}{/"};
-
-		my ( $page, $section, $ident );
-
-		# check for link patterns
-		if ( $par =~ m{^([^/]+?)/(?!")(.*?)$} ) {    # name/ident
-			    # we've got a name/ident (no quotes)
-			if ( length $2 ) {
-				( $page, $ident ) = ( $1, $2 );
-			} else {
-				( $page, $section ) = ( $1, $2 );
-			}
-			### print STDERR "--> L<$par> to page $page, ident $ident\n";
-
-		} elsif ( $par =~ m{^(.*?)/"?(.*?)"?$} ) {    # [name]/"section"
-			    # even though this should be a "section", we go for ident first
-			( $page, $ident ) = ( $1, $2 );
-			### print STDERR "--> L<$par> to page $page, section $section\n";
-
-		} elsif ( $par =~ /\s/ ) {  # this must be a section with missing quotes
-			( $page, $section ) = ( '', $par );
-			### print STDERR "--> L<$par> to void page, section $section\n";
-
-		} else {
-			( $page, $section ) = ( $par, '' );
-			### print STDERR "--> L<$par> to page $par, void section\n";
-		}
-
-		# now, either $section or $ident is defined. the convoluted logic
-		# below tries to resolve L<> according to what the user specified.
-		# failing this, we try to find the next best thing...
-		my ( $url, $ltext, $fid );
-
-	RESOLVE: {
-			if ( defined $ident ) {
-				## try to resolve $ident as an item
-				( $url, $fid ) = coderef( $page, $ident );
-				if ($url) {
-					if ( !defined($linktext) ) {
-						$linktext = $ident;
-						$linktext .= " in " if $ident && $page;
-						$linktext .= "the $page manpage" if $page;
-					}
-					###  print STDERR "got coderef url=$url\n";
-					last RESOLVE;
-				}
-				## no luck: go for a section (auto-quoting!)
-				$section = $ident;
-			}
-			## now go for a section
-			my $htmlsection = htmlify($section);
-			$url = page_sect( $page, $htmlsection );
-			if ($url) {
-				if ( !defined($linktext) ) {
-					$linktext = $section;
-					$linktext .= " in " if $section && $page;
-					$linktext .= "the $page manpage" if $page;
-				}
-				### print STDERR "got page/section url=$url\n";
-				last RESOLVE;
-			}
-			## no luck: go for an ident
-			if ($section) {
-				$ident = $section;
-			} else {
-				$ident = $page;
-				$page  = undef();
-			}
-			( $url, $fid ) = coderef( $page, $ident );
-			if ($url) {
-				if ( !defined($linktext) ) {
-					$linktext = $ident;
-					$linktext .= " in " if $ident && $page;
-					$linktext .= "the $page manpage" if $page;
-				}
-				### print STDERR "got section=>coderef url=$url\n";
-				last RESOLVE;
-			}
-
-			# warning; show some text.
-			$linktext = $opar unless defined $linktext;
-			warn
-"$0: $Podfile: cannot resolve L<$opar> in paragraph $Paragraph.\n"
-				unless $Quiet;
-		}
-
-		# now we have a URL or just plain code
-		$$rstr = $linktext . '>' . $$rstr;
-		if ( defined($url) ) {
-			$res = "<a href=\"$url\">" . process_text1( $lev, $rstr ) . '</a>';
-		} else {
-			$res = '<em>' . process_text1( $lev, $rstr ) . '</em>';
-		}
-
-	} elsif ( $func eq 'S' ) {
-
-		# S<text> - non-breaking spaces
-		$res = process_text1( $lev, $rstr );
-		$res =~ s/ /&nbsp;/g;
-
-	} elsif ( $func eq 'X' ) {
-
-		# X<> - ignore
-		warn "$0: $Podfile: invalid X<> in paragraph $Paragraph.\n"
-			unless $$rstr =~ s/^[^>]*>//
-				or $Quiet;
-	} elsif ( $func eq 'Z' ) {
-
-		# Z<> - empty
-		warn "$0: $Podfile: invalid Z<> in paragraph $Paragraph.\n"
-			unless $$rstr =~ s/^>//
-				or $Quiet;
-
-	} else {
-		my $term = pattern $closing;
-		while ( $$rstr =~ s/\A(.*?)(([BCEFILSXZ])<(<+[^\S\n]+)?|$term)//s ) {
-
-			# all others: either recurse into new function or
-			# terminate at closing angle bracket(s)
-			my $pt = $1;
-			$pt .= $2 if !$3 && $lev == 1;
-			$res .= $lev == 1 ? pure_text($pt) : inIS_text($pt);
-			return $res if !$3 && $lev > 1;
-			if ($3) {
-				$res .= process_text1( $lev, $rstr, $3, closing $4 );
-			}
-		}
-		if ( $lev == 1 ) {
-			$res .= pure_text($$rstr);
-		} elsif ( !$Quiet ) {
-			my $snippet = substr( $$rstr, 0, 60 );
-			warn
-"$0: $Podfile: undelimited $func<> in paragraph $Paragraph: '$snippet'.\n"
-
-		}
-		$res = process_text_rfc_links($res);
-	}
-	return $res;
-}
-
-#
-# go_ahead: extract text of an IS (can be nested)
-#
-sub go_ahead($$$) {
-	my ( $rstr, $func, $closing ) = @_;
-	my $res     = '';
-	my @closing = ($closing);
-	while ( $$rstr =~
-		s/\A(.*?)(([BCEFILSXZ])<(<+\s+)?|@{[pattern $closing[0]]})//s )
-	{
-		$res .= $1;
-		unless ($3) {
-			shift @closing;
-			return $res unless @closing;
-		} else {
-			unshift @closing, closing $4;
-		}
-		$res .= $2;
-	}
-	unless ($Quiet) {
-		my $snippet = substr( $$rstr, 0, 60 );
-		warn
-"$0: $Podfile: undelimited $func<> in paragraph $Paragraph (go_ahead): '$snippet'.\n";
-	}
-	return $res;
-}
-
-#
-# emit_C - output result of C<text>
-#    $text is the depod-ed text
-#
-sub emit_C($;$$) {
-	my ( $text, $nocode, $args ) = @_;
-	$args = '' unless defined $args;
-	my $res;
-	my ( $url, $fid ) = coderef( undef(), $text );
-
-	# need HTML-safe text
-	my $linktext = html_escape("$text$args");
-
-	if ( $text !~ /^[\$@%]/
-		&& defined($url)
-		&& ( !defined($EmittedItem) || $EmittedItem ne $fid ) )
-	{
-		$res = "<a href=\"$url\"><code>$linktext</code></a>";
-	} elsif ( 0 && $nocode ) {
-		$res = $linktext;
-	} else {
-		$res = "<code>$linktext</code>";
-	}
-	return $res;
-}
-
-#
-# html_escape: make text safe for HTML
-#
-sub html_escape {
-	my $rest = $_[0];
-	$rest =~ s/&/&amp;/g;
-	$rest =~ s/</&lt;/g;
-	$rest =~ s/>/&gt;/g;
-	$rest =~ s/"/&quot;/g;
-
-	# &apos; is only in XHTML, not HTML4.  Be conservative
-	#$rest   =~ s/'/&apos;/g;
-	return $rest;
-}
-
-#
-# dosify - convert filenames to 8.3
-#
-sub dosify {
-	my ($str) = @_;
-	return lc($str) if $^O eq 'VMS';    # VMS just needs casing
-	if ($Is83) {
-		$str = lc $str;
-		$str =~ s/(\.\w+)/substr ($1,0,4)/ge;
-		$str =~ s/(\w+)/substr ($1,0,8)/ge;
-	}
-	return $str;
-}
-
-#
-# page_sect - make a URL from the text of a L<>
-#
-sub page_sect($$) {
-	my ( $page, $section ) = @_;
-	my ( $linktext, $page83, $link );    # work strings
-
-	# check if we know that this is a section in this page
-	if ( !defined $Pages{$page} && defined $Sections{$page} ) {
-		$section = $page;
-		$page    = "";
-		### print STDERR "reset page='', section=$section\n";
-	}
-
-	$page83 = dosify($page);
-	$page = $page83 if ( defined $Pages{$page83} );
-	if ( $page eq "" ) {
-		$link = "#" . anchorify($section);
-	} elsif ( $page =~ /::/ ) {
-		$page =~ s,::,/,g;
-
-		# Search page cache for an entry keyed under the html page name,
-		# then look to see what directory that page might be in.  NOTE:
-		# this will only find one page. A better solution might be to produce
-		# an intermediate page that is an index to all such pages.
-		my $page_name = $page;
-		$page_name =~ s,^.*/,,s;
-		if ( defined( $Pages{$page_name} )
-			&& $Pages{$page_name} =~ /([^:]*$page)\.(?:pod|pm):/ )
-		{
-			$page = $1;
-		} else {
-
-			# NOTE: This branch assumes that all A::B pages are located in
-			# $Htmlroot/A/B.html . This is often incorrect, since they are
-			# often in $Htmlroot/lib/A/B.html or such like. Perhaps we could
-			# analyze the contents of %Pages and figure out where any
-			# cousins of A::B are, then assume that.  So, if A::B isn't found,
-			# but A::C is found in lib/A/C.pm, then A::B is assumed to be in
-			# lib/A/B.pm. This is also limited, but it's an improvement.
-			# Maybe a hints file so that the links point to the correct places
-			# nonetheless?
-
-		}
-		$link = "$Htmlroot/$page.html";
-		$link .= "#" . anchorify($section) if ($section);
-	} elsif ( !defined $Pages{$page} ) {
-		$link = "";
-	} else {
-		$section = anchorify($section) if $section ne "";
-		### print STDERR "...section=$section\n";
-
-		# if there is a directory by the name of the page, then assume that an
-		# appropriate section will exist in the subdirectory
-		#	if ($section ne "" && $Pages{$page} =~ /([^:]*[^(\.pod|\.pm)]):/) {
-		if ( $section ne "" && $Pages{$page} =~ /([^:]*(?<!\.pod)(?<!\.pm)):/ )
-		{
-			$link = "$Htmlroot/$1/$section.html";
-			### print STDERR "...link=$link\n";
-
-		 # since there is no directory by the name of the page, the section will
-		 # have to exist within a .html of the same name.  thus, make sure there
-		 # is a .pod or .pm that might become that .html
-		} else {
-			$section = "#$section" if $section;
-			### print STDERR "...section=$section\n";
-
-			# check if there is a .pod with the page name.
-			# for L<Foo>, Foo.(pod|pm) is preferred to A/Foo.(pod|pm)
-			if ( $Pages{$page} =~ /([^:]*)\.(?:pod|pm):/ ) {
-				$link = "$Htmlroot/$1.html$section";
-			} else {
-				$link = "";
-			}
-		}
-	}
-
-	if ($link) {
-
-		# Here, we take advantage of the knowledge that $Htmlfileurl ne ''
-		# implies $Htmlroot eq ''. This means that the link in question
-		# needs a prefix of $Htmldir if it begins with '/'. The test for
-		# the initial '/' is done to avoid '#'-only links, and to allow
-		# for other kinds of links, like file:, ftp:, etc.
-		my $url;
-		if ( $Htmlfileurl ne '' ) {
-			$link = "$Htmldir$link" if $link =~ m{^/}s;
-			$url = relativize_url( $link, $Htmlfileurl );
-
-			# print( "  b: [$link,$Htmlfileurl,$url]\n" );
-		} else {
-			$url = $link;
-		}
-		return $url;
-
-	} else {
-		return undef();
-	}
-}
-
-#
-# relativize_url - convert an absolute URL to one relative to a base URL.
-# Assumes both end in a filename.
-#
-sub relativize_url {
-	my ( $dest, $source ) = @_;
-
-	my ( $dest_volume, $dest_directory, $dest_file ) =
-		File::Spec::Unix->splitpath($dest);
-	$dest = File::Spec::Unix->catpath( $dest_volume, $dest_directory, '' );
-
-	my ( $source_volume, $source_directory, $source_file ) =
-		File::Spec::Unix->splitpath($source);
-	$source =
-		File::Spec::Unix->catpath( $source_volume, $source_directory, '' );
-
-	my $rel_path = '';
-	if ( $dest ne '' ) {
-		$rel_path = File::Spec::Unix->abs2rel( $dest, $source );
-	}
-
-	if (   $rel_path ne ''
-		&& substr( $rel_path, -1 ) ne '/'
-		&& substr( $dest_file, 0, 1 ) ne '#' )
-	{
-		$rel_path .= "/$dest_file";
-	} else {
-		$rel_path .= "$dest_file";
-	}
-
-	return $rel_path;
-}
-
-#
-# coderef - make URL from the text of a C<>
-#
-sub coderef($$) {
-	my ( $page, $item ) = @_;
-	my ($url);
-
-	my $fid = fragment_id($item);
-
-	if ( defined($page) && $page ne "" ) {
-
-		# we have been given a $page...
-		$page =~ s{::}{/}g;
-
-		Carp::confess(
-"Undefined fragment '$item' from fragment_id() in coderef() in $Podfile"
-		) if !defined $fid;
-
-		# Do we take it? Item could be a section!
-		my $base = $Items{$fid} || "";
-		$base =~ s{[^/]*/}{};
-		if ( $base ne "$page.html" ) {
-			###   print STDERR "coderef( $page, $item ): items{$fid} = $Items{$fid} = $base => discard page!\n";
-			$page = undef();
-		}
-
-	} else {
-
-		# no page - local items precede cached items
-		if ( defined($fid) ) {
-			if ( exists $Local_Items{$fid} ) {
-				$page = $Local_Items{$fid};
-			} else {
-				$page = $Items{$fid};
-			}
-		}
-	}
-
-	# if there was a pod file that we found earlier with an appropriate
-	# =item directive, then create a link to that page.
-	if ( defined $page ) {
-		if ($page) {
-			if ( exists $Pages{$page} and $Pages{$page} =~ /([^:.]*)\.[^:]*:/ )
-			{
-				$page = $1 . '.html';
-			}
-			my $link = "$Htmlroot/$page#" . anchorify($fid);
-
-			# Here, we take advantage of the knowledge that $Htmlfileurl
-			# ne '' implies $Htmlroot eq ''.
-			if ( $Htmlfileurl ne '' ) {
-				$link = "$Htmldir$link";
-				$url = relativize_url( $link, $Htmlfileurl );
-			} else {
-				$url = $link;
-			}
-		} else {
-			$url = "#" . anchorify($fid);
-		}
-
-		confess "url has space: $url" if $url =~ /"[^"]*\s[^"]*"/;
-	}
-	return ( $url, $fid );
-}
-
-#
-# Adapted from Nick Ing-Simmons' PodToHtml package.
-sub relative_url {
-	my $source_file      = shift;
-	my $destination_file = shift;
-
-	my $source = URI::file->new_abs($source_file);
-	my $uo = URI::file->new( $destination_file, $source )->abs;
-	return $uo->rel->as_string;
-}
-
-#
-# finish_list - finish off any pending HTML lists.  this should be called
-# after the entire pod file has been read and converted.
-#
-sub finish_list {
-	while ( $Listlevel > 0 ) {
-		print HTML "</dl>\n";
-		$Listlevel--;
-	}
-}
-
-#
-# htmlify - converts a pod section specification to a suitable section
-# specification for HTML. Note that we keep spaces and special characters
-# except ", ? (Netscape problem) and the hyphen (writer's problem...).
-#
-sub htmlify {
-	my ($heading) = @_;
-	$heading =~ s/(\s+)/ /g;
-	$heading =~ s/\s+\Z//;
-	$heading =~ s/\A\s+//;
-
-	# The hyphen is a disgrace to the English language.
-	# $heading =~ s/[-"?]//g;
-	$heading =~ s/["?]//g;
-	$heading = lc($heading);
-	return $heading;
-}
-
-#
-# similar to htmlify, but turns non-alphanumerics into underscores
-#
-sub anchorify {
-	my ($anchor) = @_;
-	$anchor =~ s/\([^)]*\)//;
-	$anchor = htmlify($anchor);
-	$anchor =~ s/\W/_/g;
-	$anchor =~ tr/_/_/s;
-	return $anchor;
-}
-
-#
-# depod - convert text by eliminating all interior sequences
-# Note: can be called with copy or modify semantics
-#
-my %E2c;
-$E2c{lt}     = '<';
-$E2c{gt}     = '>';
-$E2c{sol}    = '/';
-$E2c{verbar} = '|';
-$E2c{amp}    = '&';    # in Tk's pods
-
-sub depod1($;$$);
-
-sub depod($) {
-	my $string;
-	if ( ref( $_[0] ) ) {
-		$string = ${ $_[0] };
-		${ $_[0] } = depod1( \$string );
-	} else {
-		$string = $_[0];
-		depod1( \$string );
-	}
-}
-
-sub depod1($;$$) {
-	my ( $rstr, $func, $closing ) = @_;
-	my $res = '';
-	return $res unless defined $$rstr;
-	if ( !defined($func) ) {
-
-		# skip to next begin of an interior sequence
-		while ( $$rstr =~ s/\A(.*?)([BCEFILSXZ])<(<+[^\S\n]+)?//s ) {
-
-			# recurse into its text
-			$res .= $1 . depod1( $rstr, $2, closing $3);
-		}
-		$res .= $$rstr;
-	} elsif ( $func eq 'E' ) {
-
-		# E<x> - convert to character
-		$$rstr =~ s/^([^>]*)>//;
-		$res .= $E2c{$1} || "";
-	} elsif ( $func eq 'X' ) {
-
-		# X<> - ignore
-		$$rstr =~ s/^[^>]*>//;
-	} elsif ( $func eq 'Z' ) {
-
-		# Z<> - empty
-		$$rstr =~ s/^>//;
-	} else {
-
-		# all others: either recurse into new function or
-		# terminate at closing angle bracket
-		my $term = pattern $closing;
-		while ( $$rstr =~ s/\A(.*?)(([BCEFILSXZ])<(<+[^\S\n]+)?|$term)//s ) {
-			$res .= $1;
-			last unless $3;
-			$res .= depod1( $rstr, $3, closing $4 );
-		}
-		## If we're here and $2 ne '>': undelimited interior sequence.
-		## Ignored, as this is called without proper indication of where we are.
-		## Rely on process_text to produce diagnostics.
-	}
-	return $res;
-}
-
-{
-	my %seen;    # static fragment record hash
-
-	sub fragment_id_readable {
-		my $text     = shift;
-		my $generate = shift;    # optional flag
-
-		my $orig = $text;
-
-		# leave the words for the fragment identifier,
-		# change everything else to underbars.
-		$text =~
-			s/[^A-Za-z0-9_]+/_/g;    # do not use \W to avoid locale dependency.
-		$text =~ s/_{2,}/_/g;
-		$text =~ s/\A_//;
-		$text =~ s/_\Z//;
-
-		unless ($text) {
-
-			# Nothing left after removing punctuation, so leave it as is
-			# E.g. if option is named: "=item -#"
-
-			$text = $orig;
-		}
-
-		if ($generate) {
-			if ( exists $seen{$text} ) {
-
-				# This already exists, make it unique
-				$seen{$text}++;
-				$text = $text . $seen{$text};
-			} else {
-				$seen{$text} = 1;    # first time seen this fragment
-			}
-		}
-
-		$text;
-	}
-}
-
-my @HC;
-
-sub fragment_id_obfuscated {         # This was the old "_2d_2d__"
-	my $text     = shift;
-	my $generate = shift;            # optional flag
-
-	# text? Normalize by obfuscating the fragment id to make it unique
-	$text =~ s/\s+/_/sg;
-
-	$text =~ s{(\W)}{
-        defined( $HC[ord($1)] ) ? $HC[ord($1)]
-        : ( $HC[ord($1)] = sprintf( "%%%02X", ord($1) ) ) }gxe;
-	$text = substr( $text, 0, 50 );
-
-	$text;
-}
-
-#
-# fragment_id - construct a fragment identifier from:
-#   a) =item text
-#   b) contents of C<...>
-#
-
-sub fragment_id {
-	my $text     = shift;
-	my $generate = shift;    # optional flag
-
-	$text =~ s/\s+\Z//s;
-	if ($text) {
-
-		# a method or function?
-		return $1 if $text =~ /(\w+)\s*\(/;
-		return $1 if $text =~ /->\s*(\w+)\s*\(?/;
-
-		# a variable name?
-		return $1 if $text =~ /^([\$\@%*]\S+)/;
-
-		# some pattern matching operator?
-		return $1 if $text =~ m|^(\w+/).*/\w*$|;
-
-		# fancy stuff... like "do { }"
-		return $1 if $text =~ m|^(\w+)\s*{.*}$|;
-
-		# honour the perlfunc manpage: func [PAR[,[ ]PAR]...]
-		# and some funnies with ... Module ...
-		return $1 if $text =~ m{^([a-z\d_]+)(\s+[A-Z,/& ][A-Z\d,/& ]*)?$};
-		return $1 if $text =~ m{^([a-z\d]+)\s+Module(\s+[A-Z\d,/& ]+)?$};
-
-		return fragment_id_readable( $text, $generate );
-	} else {
-		return;
-	}
-}
-
-#
-# make_URL_href - generate HTML href from URL
-# Special treatment for CGI queries.
-#
-sub make_URL_href($;$) {
-	my ($url) = shift;
-	my $linktext = shift || $url;
-	if ( $url !~ s{^(http:[-\w/#~:.+=&%@!]+)(\?.*)$}{<a href="$1$2">$1</a>}i ) {
-		$url = "<a href=\"$url\">$linktext</a>";
-	}
-	return $url;
-}
-
-1;