IkiWiki/Plugin/htmlscrubber.pm

   1 #!/usr/bin/perl
   2 package IkiWiki::Plugin::htmlscrubber;
   3
   4 use warnings;
   5 use strict;
   6 use IkiWiki 2.00;
   7
   8 # This regexp matches urls that are in a known safe scheme.
   9 # Feel free to use it from other plugins.
  10 our $safe_url_regexp;
  11
  12 sub import { #{{{
  13         hook(type => "getsetup", id => "htmlscrubber", call => \&getsetup);
  14         hook(type => "sanitize", id => "htmlscrubber", call => \&sanitize);
  15
  16         # Only known uri schemes are allowed to avoid all the ways of
  17         # embedding javascrpt.
  18         # List at http://en.wikipedia.org/wiki/URI_scheme
  19         my $uri_schemes=join("|", map quotemeta,
  20                 # IANA registered schemes
  21                 "http", "https", "ftp", "mailto", "file", "telnet", "gopher",
  22                 "aaa", "aaas", "acap",  "cap", "cid", "crid",
  23                 "dav", "dict", "dns", "fax", "go", "h323", "im", "imap",
  24                 "ldap", "mid", "news", "nfs", "nntp", "pop", "pres",
  25                 "sip", "sips", "snmp", "tel", "urn", "wais", "xmpp",
  26                 "z39.50r", "z39.50s",
  27                 # Selected unofficial schemes
  28                 "aim", "callto", "cvs", "ed2k", "feed", "fish", "gg",
  29                 "irc", "ircs", "lastfm", "ldaps", "magnet", "mms",
  30                 "msnim", "notes", "rsync", "secondlife", "skype", "ssh",
  31                 "sftp", "smb", "sms", "snews", "webcal", "ymsgr",
  32         );
  33         # data is a special case. Allow data:image/*, but
  34         # disallow data:text/javascript and everything else.
  35         $safe_url_regexp=qr/^(?:(?:$uri_schemes):|data:image\/|[^:]+(?:$|\/))/i;
  36 } # }}}
  37
  38 sub getsetup () { #{{{
  39         return
  40                 plugin => {
  41                         safe => 1,
  42                         rebuild => undef,
  43                 },
  44 } #}}}
  45
  46 sub sanitize (@) { #{{{
  47         my %params=@_;
  48         return scrubber()->scrub($params{content});
  49 } # }}}
  50
  51 my $_scrubber;
  52 sub scrubber { #{{{
  53         return $_scrubber if defined $_scrubber;
  54
  55         eval q{use HTML::Scrubber};
  56         error($@) if $@;
  57         # Lists based on http://feedparser.org/docs/html-sanitization.html
  58         # With html 5 video and audio tags added.
  59         $_scrubber = HTML::Scrubber->new(
  60                 allow => [qw{
  61                         a abbr acronym address area b big blockquote br br/
  62                         button caption center cite code col colgroup dd del
  63                         dfn dir div dl dt em fieldset font form h1 h2 h3 h4
  64                         h5 h6 hr hr/ i img input ins kbd label legend li map
  65                         menu ol optgroup option p p/ pre q s samp select small
  66                         span strike strong sub sup table tbody td textarea
  67                         tfoot th thead tr tt u ul var
  68                         video audio
  69                 }],
  70                 default => [undef, { (
  71                         map { $_ => 1 } qw{
  72                                 abbr accept accept-charset accesskey
  73                                 align alt axis border cellpadding cellspacing
  74                                 char charoff charset checked class
  75                                 clear cols colspan color compact coords
  76                                 datetime dir disabled enctype for frame
  77                                 headers height hreflang hspace id ismap
  78                                 label lang maxlength media method
  79                                 multiple name nohref noshade nowrap prompt
  80                                 readonly rel rev rows rowspan rules scope
  81                                 selected shape size span start summary
  82                                 tabindex target title type valign
  83                                 value vspace width
  84                                 autoplay loopstart loopend end
  85                                 playcount controls
  86                         } ),
  87                         "/" => 1, # emit proper <hr /> XHTML
  88                         href => $safe_url_regexp,
  89                         src => $safe_url_regexp,
  90                         action => $safe_url_regexp,
  91                         cite => $safe_url_regexp,
  92                         longdesc => $safe_url_regexp,
  93                         poster => $safe_url_regexp,
  94                         usemap => $safe_url_regexp,
  95                 }],
  96         );
  97         return $_scrubber;
  98 } # }}}
  99
 100 1