]> sipb.mit.edu Git - ikiwiki.git/blobdiff - IkiWiki/Plugin/htmlscrubber.pm
export $safe_url_regexp
[ikiwiki.git] / IkiWiki / Plugin / htmlscrubber.pm
index 489827167101ebcd2445d6936a8a80a065086adb..8c37f1be618b612c68bab5350077d341668be4d6 100644 (file)
@@ -5,8 +5,33 @@ use warnings;
 use strict;
 use IkiWiki 2.00;
 
 use strict;
 use IkiWiki 2.00;
 
+# This regexp matches urls that are in a known safe scheme.
+# Feel free to use it from other plugins.
+our $safe_url_regexp;
+
 sub import { #{{{
        hook(type => "sanitize", id => "htmlscrubber", call => \&sanitize);
 sub import { #{{{
        hook(type => "sanitize", id => "htmlscrubber", call => \&sanitize);
+
+       # Only known uri schemes are allowed to avoid all the ways of
+       # embedding javascrpt.
+       # List at http://en.wikipedia.org/wiki/URI_scheme
+       my $uri_schemes=join("|",
+               # IANA registered schemes
+               "http", "https", "ftp", "mailto", "file", "telnet", "gopher",
+               "aaa", "aaas", "acap",  "cap", "cid", "crid", 
+               "dav", "dict", "dns", "fax", "go", "h323", "im", "imap",
+               "ldap", "mid", "news", "nfs", "nntp", "pop", "pres",
+               "sip", "sips", "snmp", "tel", "urn", "wais", "xmpp",
+               "z39.50r", "z39.50s",
+               # Selected unofficial schemes
+               "aim", "callto", "cvs", "ed2k", "feed", "fish", "gg",
+               "irc", "ircs", "lastfm", "ldaps", "magnet", "mms",
+               "msnim", "notes", "rsync", "secondlife", "skype", "ssh",
+               "sftp", "sms", "steam", "webcal", "ymsgr",
+       );
+       # data is a special case. Allow data:image/*, but
+       # disallow data:text/javascript and everything else.
+       $safe_url_regexp=qr/^(?:(?:$uri_schemes):|data:image\/|[^:]+$)/i;
 } # }}}
 
 sub sanitize (@) { #{{{
 } # }}}
 
 sub sanitize (@) { #{{{
@@ -17,40 +42,48 @@ sub sanitize (@) { #{{{
 my $_scrubber;
 sub scrubber { #{{{
        return $_scrubber if defined $_scrubber;
 my $_scrubber;
 sub scrubber { #{{{
        return $_scrubber if defined $_scrubber;
-       
+
        eval q{use HTML::Scrubber};
        error($@) if $@;
        # Lists based on http://feedparser.org/docs/html-sanitization.html
        eval q{use HTML::Scrubber};
        error($@) if $@;
        # Lists based on http://feedparser.org/docs/html-sanitization.html
+       # With html 5 video and audio tags added.
        $_scrubber = HTML::Scrubber->new(
                allow => [qw{
        $_scrubber = HTML::Scrubber->new(
                allow => [qw{
-                       a abbr acronym address area b big blockquote br
+                       a abbr acronym address area b big blockquote br br/
                        button caption center cite code col colgroup dd del
                        dfn dir div dl dt em fieldset font form h1 h2 h3 h4
                        button caption center cite code col colgroup dd del
                        dfn dir div dl dt em fieldset font form h1 h2 h3 h4
-                       h5 h6 hr i img input ins kbd label legend li map
-                       menu ol optgroup option p pre q s samp select small
+                       h5 h6 hr hr/ i img input ins kbd label legend li map
+                       menu ol optgroup option p p/ pre q s samp select small
                        span strike strong sub sup table tbody td textarea
                        tfoot th thead tr tt u ul var
                        span strike strong sub sup table tbody td textarea
                        tfoot th thead tr tt u ul var
+                       video audio
                }],
                default => [undef, { (
                        map { $_ => 1 } qw{
                }],
                default => [undef, { (
                        map { $_ => 1 } qw{
-                               abbr accept accept-charset accesskey action
+                               abbr accept accept-charset accesskey
                                align alt axis border cellpadding cellspacing
                                align alt axis border cellpadding cellspacing
-                               char charoff charset checked cite class
+                               char charoff charset checked class
                                clear cols colspan color compact coords
                                datetime dir disabled enctype for frame
                                clear cols colspan color compact coords
                                datetime dir disabled enctype for frame
-                               headers height href hreflang hspace id ismap
-                               label lang longdesc maxlength media method
+                               headers height hreflang hspace id ismap
+                               label lang maxlength media method
                                multiple name nohref noshade nowrap prompt
                                readonly rel rev rows rowspan rules scope
                                multiple name nohref noshade nowrap prompt
                                readonly rel rev rows rowspan rules scope
-                               selected shape size span src start summary
-                               tabindex target title type usemap valign
+                               selected shape size span start summary
+                               tabindex target title type valign
                                value vspace width
                                value vspace width
+                               autoplay loopstart loopend end
+                               playcount controls 
                        } ),
                        "/" => 1, # emit proper <hr /> XHTML
                        } ),
                        "/" => 1, # emit proper <hr /> XHTML
-                       "style" => qr{^[-a-zA-Z0-9]+$}, # only very simple
-                                                       # references allowed,
-                                                       # to avoid javascript
-                       }],
+                       href => $safe_url_regexp,
+                       src => $safe_url_regexp,
+                       action => $safe_url_regexp,
+                       cite => $safe_url_regexp,
+                       longdesc => $safe_url_regexp,
+                       poster => $safe_url_regexp,
+                       usemap => $safe_url_regexp,
+               }],
        );
        return $_scrubber;
 } # }}}
        );
        return $_scrubber;
 } # }}}