Enhancement for AJAX sites via google loophole

[alexbobp]

Google has published a specification for allowing ajax sites to load content on a single page through javascript, but still be crawlable. Basically there is a rewrite rule you can apply to fragment links to get an actual http link to the content.

https://code.google.com/web/ajaxcrawlin ... arted.html

Lots of sites (twitter for example) seem to be supporting this. If noscript adds this feature, it could let us browse such sites with javascript turned off.

[Giorgio Maone]

Very interesting, thank you.

[Giorgio Maone]

Please check latest development build.
See http://lifehacker.com/#!5465246/mozilla ... an-malware for a working example.

[dhouwn]

See http://lifehacker.com/#!5465246/mozilla ... an-malware for a working example.

Apparently not working with the current 4.0 beta:

Code: Select all

[NoScript XSS] xss.reason.ReferenceError: qs is not defined --- ([object XPCWrappedNative_NoHelper],"http-on-modify-request",null)@chrome://noscript/content/RequestWatchdog.js:84
loadURIWithFlags("http://lifehacker.com/#!5465246/mozilla-warns-of-two-firefox-extensions-containing-trojan-malware",0,[object XPCWrappedNative_NoHelper],"UTF-8",(void 0))@chrome://global/content/bindings/browser.xml:193
addTab("http://lifehacker.com/#!5465246/mozilla-warns-of-two-firefox-extensions-containing-trojan-malware",[object XPCWrappedNative_NoHelper])@chrome://browser/content/tabbrowser.xml:1326
loadOneTab("http://lifehacker.com/#!5465246/mozilla-warns-of-two-firefox-extensions-containing-trojan-malware",[object XPCWrappedNative_NoHelper])@chrome://browser/content/tabbrowser.xml:1093
openLinkIn("http://lifehacker.com/#!5465246/mozilla-warns-of-two-firefox-extensions-containing-trojan-malware","tab",[object Object])@chrome://browser/content/utilityOverlay.js:237
handleLinkClick([object MouseEvent],"http://lifehacker.com/#!5465246/mozilla-warns-of-two-firefox-extensions-containing-trojan-malware",[object Proxy])@chrome://browser/content/browser.js:10498
contentAreaClick([object MouseEvent],false)@chrome://browser/content/browser.js:10461
onclick([object MouseEvent])@chrome://browser/content/browser.xul:1

[al_9x]

Apparently not working with the current 4.0 beta:

neither in 3.6.13

[Giorgio Maone]

Yep, there's a bug which slipped in 10 seconds before release and I'm fixing right now.

[Giorgio Maone]

Fixed in latest development build

[alexbobp]

Awesome. I'm glad you liked the idea, and I'm amazed you implemented it so fast.
Lifehacker article permalinks work, and that's pretty excellent. (The homepage is still broken though, I guess it uses javascript to populate it's frame)

Excellent work! I'll be pitching a few bucks towards your beer fund.

[al_9x]

On this new style ajax site, the next post click fails when doubleclick.net is blocked. The quickest way to repro is on gawker.com.

Fx 3.6.13, NS 2.0.9.9rc2, new profile

1. go to http://gawker.com/
2. allow gawker.com, gawkerassets.com
3. after refresh, the content should appear, click "NEXT POST" (bottom left corner)
4. the ajax nav never fully completes, a couple of js errors are logged:

   Code: Select all

   Error: too much recursion
   Source File: http://fastcache.gawkerassets.com/assets/base.v10/static/base.v10.jqueryplugin.20110217.js
   Line: 6
   Error: too much recursion
   Source File: http://fastcache.gawkerassets.com/assets/base.v10/static/base.v10.jquery.20110217.js
   Line: 16
   

5. If you allow doubleclick.net at gawker.com, "NEXT POST" ajax nav completes

[Giorgio Maone]

It actually works for me. Maybe they already fixed it?

[al_9x]

It actually works for me. Maybe they already fixed it?

No, still happening with 100% consistency on a new profile, the content loads, but remains covered by a semi-transparent overlay with a spinner.

[dhouwn]

Can't reproduce, "NEXT POST" works fine, some exceptions like the following get logged though:

Code: Select all

Error: jQuery(".t_" + d.thread_id).html(d.content) is null
Source File: http://fastcache.gawkerassets.com/assets/base.v10/static/base.v10.widget.20110217.js
Line: 42

BTW, I have subdomain-matching turned on and just allowed gawker.com and fastcache.gawkerassets.com.

[al_9x]

They changed their interface, there's no longer a next post button, but a list of posts. I can still repro by clicking on any post title. 2 "too much recursion" errors and it stays in this state:


To resolve, you have to allow doubleclick.net and also the script that doubleclick in turn loads 2mdn.net. It used to take just doubleclick. That you can't repro is an indication that there probably isn't a real dependency on doubleclick but something timing sensitive, can you perhaps try on your slower laptop?

[therube]

2. allow gawker.com, gawkerassets.com

Then I allowed an Object. *@http://gawker.com, I think it was.

After that, the links worked as expected.

After a Reset, a manual refresh may be needed.

Looks like sometimes a link will load fully, sometimes partially, sometimes opaque like your screenshot.
Also looks like they have an object which nukeanythingenhanced can zap to clear some "opaque" up.

(The double vertical scroll is totally weird.)

[al_9x]

Looks like sometimes a link will load fully, sometimes partially, sometimes opaque like your screenshot.

Well, it looks like you can repro intermittently, for me it's consistent. When it's like in my screenshot, do you get the "too much recursion" errors in the console?