http://www.google.com/url?q=%s). '.
'This prevents PageRank leaks, while being as transparent as possible '.
'to users (you may also want to add some client side JavaScript to '.
'override the text in the statusbar). Warning: many security experts '.
'believe that this form of protection does not deter spam-bots. '.
'You can also use this directive to redirect users to a splash page '.
'telling them they are leaving your website. '.
'This directive has been available since 1.3.0.'
);
/**
* Validates a URI as defined by RFC 3986.
* @note Scheme-specific mechanics deferred to HTMLPurifier_URIScheme
*/
class HTMLPurifier_AttrDef_URI extends HTMLPurifier_AttrDef
{
var $host;
var $PercentEncoder;
var $embeds_resource;
/**
* @param $embeds_resource_resource Does the URI here result in an extra HTTP request?
*/
function HTMLPurifier_AttrDef_URI($embeds_resource = false) {
$this->host = new HTMLPurifier_AttrDef_Host();
$this->PercentEncoder = new HTMLPurifier_PercentEncoder();
$this->embeds_resource = (bool) $embeds_resource;
}
function validate($uri, $config, &$context) {
// We'll write stack-based parsers later, for now, use regexps to
// get things working as fast as possible (irony)
// parse as CDATA
$uri = $this->parseCDATA($uri);
// fix up percent-encoding
$uri = $this->PercentEncoder->normalize($uri);
// while it would be nice to use parse_url(), that's specifically
// for HTTP and thus won't work for our generic URI parsing
// according to the RFC... (but this cuts corners, i.e. non-validating)
$r_URI = '!'.
'(([^:/?#<>\'"]+):)?'. // 2. Scheme
'(//([^/?#<>\'"]*))?'. // 4. Authority
'([^?#<>\'"]*)'. // 5. Path
'(\?([^#<>\'"]*))?'. // 7. Query
'(#([^<>\'"]*))?'. // 8. Fragment
'!';
$matches = array();
$result = preg_match($r_URI, $uri, $matches);
if (!$result) return false; // invalid URI
// seperate out parts
$scheme = !empty($matches[1]) ? $matches[2] : null;
$authority = !empty($matches[3]) ? $matches[4] : null;
$path = $matches[5]; // always present, can be empty
$query = !empty($matches[6]) ? $matches[7] : null;
$fragment = !empty($matches[8]) ? $matches[9] : null;
$registry =& HTMLPurifier_URISchemeRegistry::instance();
if ($scheme !== null) {
// no need to validate the scheme's fmt since we do that when we
// retrieve the specific scheme object from the registry
$scheme = ctype_lower($scheme) ? $scheme : strtolower($scheme);
$scheme_obj =& $registry->getScheme($scheme, $config, $context);
if (!$scheme_obj) return false; // invalid scheme, clean it out
} else {
$scheme_obj =& $registry->getScheme(
$config->get('URI', 'DefaultScheme'), $config, $context
);
}
// the URI we're processing embeds_resource a resource in the page, but the URI
// it references cannot be located
if ($this->embeds_resource && !$scheme_obj->browsable) {
return false;
}
if ($authority !== null) {
// remove URI if it's absolute and we disabled externals or
// if it's absolute and embedded and we disabled external resources
unset($our_host);
if (
$config->get('URI', 'DisableExternal') ||
(
$config->get('URI', 'DisableExternalResources') &&
$this->embeds_resource
)
) {
$our_host = $config->get('URI', 'Host');
if ($our_host === null) return false;
}
$HEXDIG = '[A-Fa-f0-9]';
$unreserved = 'A-Za-z0-9-._~'; // make sure you wrap with []
$sub_delims = '!$&\'()'; // needs []
$pct_encoded = "%$HEXDIG$HEXDIG";
$r_userinfo = "(?:[$unreserved$sub_delims:]|$pct_encoded)*";
$r_authority = "/^(($r_userinfo)@)?(\[[^\]]+\]|[^:]*)(:(\d*))?/";
$matches = array();
preg_match($r_authority, $authority, $matches);
// overloads regexp!
$userinfo = !empty($matches[1]) ? $matches[2] : null;
$host = !empty($matches[3]) ? $matches[3] : null;
$port = !empty($matches[4]) ? $matches[5] : null;
// validate port
if ($port !== null) {
$port = (int) $port;
if ($port < 1 || $port > 65535) $port = null;
}
$host = $this->host->validate($host, $config, $context);
if ($host === false) $host = null;
// more lenient absolute checking
if (isset($our_host)) {
$host_parts = array_reverse(explode('.', $host));
// could be cached
$our_host_parts = array_reverse(explode('.', $our_host));
foreach ($our_host_parts as $i => $discard) {
if (!isset($host_parts[$i])) return false;
if ($host_parts[$i] != $our_host_parts[$i]) return false;
}
}
// userinfo and host are validated within the regexp
} else {
$port = $host = $userinfo = null;
}
// query and fragment are quite simple in terms of definition:
// *( pchar / "/" / "?" ), so define their validation routines
// when we start fixing percent encoding
// path gets to be validated against a hodge-podge of rules depending
// on the status of authority and scheme, but it's not that important,
// esp. since it won't be applicable to everyone
// okay, now we defer execution to the subobject for more processing
// note that $fragment is omitted
list($userinfo, $host, $port, $path, $query) =
$scheme_obj->validateComponents(
$userinfo, $host, $port, $path, $query, $config, $context
);
// reconstruct authority
$authority = null;
if (!is_null($userinfo) || !is_null($host) || !is_null($port)) {
$authority = '';
if($userinfo !== null) $authority .= $userinfo . '@';
$authority .= $host;
if($port !== null) $authority .= ':' . $port;
}
// reconstruct the result
$result = '';
if ($scheme !== null) $result .= "$scheme:";
if ($authority !== null) $result .= "//$authority";
$result .= $path;
if ($query !== null) $result .= "?$query";
if ($fragment !== null) $result .= "#$fragment";
// munge if necessary
$munge = $config->get('URI', 'Munge');
if (!empty($scheme_obj->browsable) && $munge !== null) {
if ($authority !== null) {
$result = str_replace('%s', rawurlencode($result), $munge);
}
}
return $result;
}
}
?>