Changing default "failed to connect" page.

Discuss how to use and promote Web standards with the Mozilla Gecko engine.
Posts: 22
Joined: October 28th, 2005, 4:31 am
Location: Land of Hope & Glory

Post Posted September 27th, 2008, 5:26 am

Hello everyone,

Is there a way to change the default system "Failed To Connect" page?

I have a web application that refreshes a web page periodicaly to check for a status set by the application. Once in a while, the internet connection may be temporarily unavailable, so the Failed To Connect page comes up. Once that happens, the application dies, because the refresh control goes with it.

I'd like to add my own page (stored on the local computer), which will be able to maintain a periodic refresh and recover the application as soon as the internet connection kicks back in.

Am I asking the earth, or is something like this possible? Clearly, the page is HTML, so I can see potential, but where do I find it? How is it generated?

Code: Select all
<html xmlns="">
    <title>Page Load Error</title>
    <link rel="stylesheet" href="chrome://global/skin/netError.css" type="text/css" media="all"/>
    <!-- If the location of the favicon is changed here, the FAVICON_ERRORPAGE_URL symbol in
         toolkit/components/places/src/nsFaviconService.h should be updated. -->
    <link rel="icon" type="image/png" id="favicon" href="chrome://global/skin/icons/warning-16.png"/>

    <script type="application/x-javascript"><![CDATA[
      // Error url MUST be formatted like this:
      //   moz-neterror:page?e=error&u=url&d=desc
      // or optionally, to specify an alternate CSS class to allow for
      // custom styling and favicon:
      //   moz-neterror:page?e=error&u=url&s=classname&d=desc

      // Note that this file uses document.documentURI to get
      // the URL (with the format from above). This is because
      // document.location.href gets the current URI off the docshell,
      // which is the URL displayed in the location bar, i.e.
      // the URI that the user attempted to load.

      function getErrorCode()
        var url = document.documentURI;
        var error =\=/);
        var duffUrl =\&u\=/);
        return decodeURIComponent(url.slice(error + 2, duffUrl));

      function getCSSClass()
        var url = document.documentURI;
        var matches = url.match(/s\=([^&]+)\&/);
        // s is optional, if no match just return nothing
        if (!matches || matches.length < 2)
          return "";

        // parenthetical match is the second entry
        return decodeURIComponent(matches[1]);

      function getDescription()
        var url = document.documentURI;
        var desc =\=/);

        // desc == -1 if not found; if so, return an empty string
        // instead of what would turn out to be portions of the URI
        if (desc == -1)
          return "";

        return decodeURIComponent(url.slice(desc + 2));

      function retryThis(buttonEl)
        // Session history has the URL of the page that failed
        // to load, not the one of the error page. So, just call
        // reload(), which will also repost POST data correctly.
        try {
        } catch (e) {
          // We probably tried to reload a URI that caused an exception to
          // occur;  e.g. a non-existent file.

        buttonEl.disabled = true;

      function initPage()
        var err = getErrorCode();
        // if it's an unknown error or there's no title or description
        // defined, get the generic message
        var errTitle = document.getElementById("et_" + err);
        var errDesc  = document.getElementById("ed_" + err);
        if (!errTitle || !errDesc)
          errTitle = document.getElementById("et_generic");
          errDesc  = document.getElementById("ed_generic");

        var title = document.getElementById("errorTitleText");
        if (title)
          title.parentNode.replaceChild(errTitle, title);
          // change id to the replaced child's id so styling works
 = "errorTitleText";

        var sd = document.getElementById("errorShortDescText");
        if (sd)
          sd.textContent = getDescription();

        var ld = document.getElementById("errorLongDesc");
        if (ld)
          ld.parentNode.replaceChild(errDesc, ld);
          // change id to the replaced child's id so styling works
 = "errorLongDesc";

        // remove undisplayed errors to avoid bug 39098
        var errContainer = document.getElementById("errorContainer");

        var className = getCSSClass();
        if (className && className != "expertBadCert") {
          // Associate a CSS class with the root of the page, if one was passed in,
          // to allow custom styling.
          // Not "expertBadCert" though, don't want to deal with the favicon
          document.documentElement.className = className;

          // Also, if they specified a CSS class, they must supply their own
          // favicon.  In order to trigger the browser to repaint though, we
          // need to remove/add the link element.
          var favicon = document.getElementById("favicon");
          var faviconParent = favicon.parentNode;
          favicon.setAttribute("href", "chrome://global/skin/icons/" + className + "_favicon.png");
        if (className == "expertBadCert") {
        if (err == "nssBadCert") {
          // Remove the "Try again" button for security exceptions, since it's
          // almost certainly useless.
          document.getElementById("errorTryAgain").style.display = "none";
          document.getElementById("errorPageContainer").setAttribute("class", "certerror");
        else {
          // Remove the override block for non-certificate errors.  CSS-hiding
          // isn't good enough here, because of bug 39098
          var secOverride = document.getElementById("securityOverrideDiv");
      function showSecuritySection() {
        // Swap link out, content in
        document.getElementById('securityOverrideContent').style.display = '';
        document.getElementById('securityOverrideLink').style.display = 'none';
      /* In the case of SSL error pages about domain mismatch, see if
         we can hyperlink the user to the correct site.  We don't want
         to do this generically since it allows MitM attacks to redirect
         users to a site under attacker control, but in certain cases
         it is safe (and helpful!) to do so.  Bug 402210
      function addDomainErrorLink() {
        // Rather than textContent, we need to treat description as HTML
        var sd = document.getElementById("errorShortDescText");
        if (sd) {
          var desc = getDescription();
          // sanitize description text - see bug 441169
          // First, find the index of the <a> tag we care about, being careful not to
          // use an over-greedy regex
          var re = /<a id="cert_domain_link" title="([^"]+)">/;
          var result = re.exec(desc);
          // Remove sd's existing children
          sd.textContent = "";

          // Everything up to the link should be text content
          sd.appendChild(document.createTextNode(desc.slice(0, result.index)));
          // Now create the link itself
          var anchorEl = document.createElement("a");
          anchorEl.setAttribute("id", "cert_domain_link");
          anchorEl.setAttribute("title", result[1]);
          // Finally, append text for anything after the closing </a>
          sd.appendChild(document.createTextNode(desc.slice(desc.indexOf("</a>") + "</a>".length)));

        var link = document.getElementById('cert_domain_link');
        if (!link)
        var okHost = link.getAttribute("title");
        var thisHost = document.location.hostname;
        var proto = document.location.protocol;

        // If okHost is a wildcard domain ("*") let's
        // use "www" instead.  "*" isn't going to
        // get anyone anywhere useful. bug 432491
        okHost = okHost.replace(/^\*\./, "www.");

        /* case #1:
         * uses an invalid security certificate.
         * The certificate is only valid for
         * Make sure to include the "." ahead of thisHost so that
         * a MitM attack on doesn't hyperlink to ""
         * We'd normally just use a RegExp here except that we lack a
         * library function to escape them properly (bug 248062), and
         * domain names are famous for having '.' characters in them,
         * which would allow spurious and possibly hostile matches.
        if (endsWith(okHost, "." + thisHost))
          link.href = proto + okHost;

        /* case #2:
         * uses an invalid security certificate.
         * The certificate is only valid for
        if (endsWith(thisHost, "." + okHost))
          link.href = proto + okHost;
      function endsWith(haystack, needle) {
        return haystack.slice(-needle.length) == needle;


  <body dir="ltr">

    <!-- ERROR ITEM CONTAINER (removed during loading to avoid bug 39098) -->

    <!-- PAGE CONTAINER (for styling purposes only) -->
    <div id="errorPageContainer">
      <!-- Error Title -->
      <div id="errorTitle">
        <h1 id="errorTitleText">Failed to Connect</h1>

      <!-- LONG CONTENT (the section most likely to require scrolling) -->
      <div id="errorLongContent">
        <!-- Short Description -->
        <div id="errorShortDesc">
          <p id="errorShortDescText">The connection was refused when attempting to contact</p>

        <!-- Long Description (Note: See netError.dtd for used XHTML tags) -->
        <div id="errorLongDesc">

<p>Though the site seems valid, the browser was unable to establish a connection.</p>


 <li>Could the site be temporarily unavailable? Try again later.</li>

 <li>Are you unable to browse other sites?  Check the computer's network connection.</li>

 <li>Is your computer or network protected by a firewall or proxy? Incorrect settings can interfere with Web browsing.</li>


        <!-- Override section - For ssl errors only.  Removed on init for other
             error types.  -->

      <!-- Retry Button -->
      <xul:button xmlns:xul="" id="errorTryAgain" label="Try Again" oncommand="retryThis(this);"/>


    - Note: It is important to run the script this way, instead of using
    - an onload handler. This is because error pages are loaded as
    - LOAD_BACKGROUND, which means that onload handlers will not be executed.

    <script type="application/x-javascript">initPage();</script>


Posts: 22
Joined: October 28th, 2005, 4:31 am
Location: Land of Hope & Glory

Post Posted September 29th, 2008, 5:41 am


Is this a bit on the advanced side for you lot?


User avatar
Posts: 37610
Joined: October 19th, 2007, 1:53 am
Location: The South Coast of N.S.W. Oz.

Post Posted September 29th, 2008, 7:06 am

Moving to Web Dev..where the coding gurus reside.
Ancient Amateur Astronomer
Win-7-HP/IntelĀ® DualCore-2.0GHz/500G HDD/4 Gig Ram/550Watt PSU/350WattUPS/Firefox-20.0-56.0.1/T-bird-
(Always choose the "Custom" Install.)


User avatar
Posts: 5926
Joined: April 11th, 2004, 10:20 am
Location: Belgium

Post Posted September 29th, 2008, 11:13 am

This isn't really about web development... It's about changing an internal Firefox page. Probably belongs in Mozilla Development or something.


User avatar
Posts: 10077
Joined: December 19th, 2004, 12:26 am
Location: Silicon Valley, CA USA

Post Posted September 29th, 2008, 1:31 pm

Can you load the "problem page" into an <iframe> so your refresher is not dependent on the problem page loading?

Posts: 22
Joined: October 28th, 2005, 4:31 am
Location: Land of Hope & Glory

Post Posted September 30th, 2008, 8:24 am

It's in an iframe already, and the page is using:

Code: Select all
<meta http-equiv="Refresh" content="300">

to make the periodic refresh. It's a very simple solution to allowing an automatic refresh.

I'm not sure how to make a repeated refresh of the iframe from the parent page, without reloading the parent page (in which case, that would be victim to the error page if the connection went).

Because of this, my idea was to change the internal FF page to keep refreshing until the page was no longer required (the connection error condition was cleared and therefore the original page restored).

I'm specifically looking for a FF solution as I'm trying to stop my customers using other browsers, especially AOL! GAH! :shock:


User avatar
Posts: 10077
Joined: December 19th, 2004, 12:26 am
Location: Silicon Valley, CA USA

Post Posted September 30th, 2008, 1:07 pm

Will the iframe refresh if you use a script in the main page to change its src? No time to code it up now, but you would use window.setInterval or window.setTimeout in the main page.

Posts: 22
Joined: October 28th, 2005, 4:31 am
Location: Land of Hope & Glory

Post Posted September 30th, 2008, 2:15 pm

I believe you've provided a suitable solution for me. From what I can understand you're saying...

The containing window can set the refresh interval, which will carry on periodically calling the iframe window, irrespective of the iframe contents, so if the internet connection is interrupted, the refreshing will still occur, with the iframe contents resuming as normal as soon as the connection is restored.

I'm going to go give it a try! Many many thanks.

Return to Web Development / Standards Evangelism

Who is online

Users browsing this forum: No registered users and 1 guest