I haven't managed to get this also running in IE8 and below yet. Basically, it should look something like

varaxContext,defaultContext=document;if(jQuery.isFunction(document.implementation.createHTMLDocument)){defaultContext=document.implementation.createHTMLDocument();}elseif("ActiveXObject"inwindow){axContext=newActiveXObject("htmlfile");axContext.write();axContext.close();defaultContext=axContext.body;}context=context||defaultContext||document;

But using the ActiveX context breaks quite a bunch of tests. But as this patch still helps for most other browsers, it's still an improvement.

Note: I'll update this PR as soon as#1505 is merged into 2.x

I think the problem with IE<=8 is that you end up with nodes from a foreign document that can't be added directly to the target document. If you were writing a plugin to fix this you could traverse and sanitize the elements in the foreign document, then serialize the sanitized tree to an HTML string and parse that in the target document.

Just to give it a ticket, this is#11974 which can be reopened.

So to be clear, the most common internal use of.parseHTML() will mean that the next step will be to append the HTML to the document,triggering the inline handlers. However, we have anopen ticket #14228 to provide a hook point for sanitizing input, and perhaps we could integrate a strategy into that.

Again, I suspect a complete solution is a bit heavy for including into core but would like to make it easy for someone to do the work in a plugin.

That's why I called the PR "mitigate" not "fix". ;) I think it's a good start to address the issue at this point. I won't put more effort in fixing this for IE<=8 – this patch improves the handling for ~90% of the users, for the rest it's simply the same as before.

A plug-in seems like a good idea – would parseHTML be the only function to hook into? It would basically use the changes proposed in#1508, but as the user has to install this separately, it won't have unwanted side effects in core.

Since this is the one of the last1.x-master PR's and it seems solution wouldn't cover all browsers but fix for#1747 will, so i would like to close it and#1505 too.

Per thecomment above I think this is actually part of the solution forgh-1747 because it prevents some inline code from running before the HTML is appended to the document. The two PRs could be merged though.

@dmethvin misinterpreted your comment, now i see the conundrum, so we have two problems:createHTMLDocument doesn't exist in IE8 and in old android otherwise we good.

SocreateHTMLDocument does exist in android 2.3 and in IE8 we could use an active X object.

Sounds good?

#1505 is for jQuery 2.x, the proposed solution works fine there.

Regarding this PR, IE8 could use anActiveX component for filtering, but it seems to break some stuff in jQuery I wasn't able to figure out.

A quick (although not complete) solution would be to usedocument.implementation.createHTMLDocument where it's available, otherwise falling back to the current behavior. Although this would leave IE8 and below still vulnerable to those kind of injections, the issue would be fixed for all other browsers.

The Active X component would require some additional work and could be added to a later patch release.

I think the quick incomplete solution would be fine. That leaves the problem unsolved for IE8, but many problems are unsolved for IE8. 😈

Fixed execution with activex component, but it was throwing in indeed unexpected places, given thatparseHTML is one is the basis for many code paths i think risks are pretty high, so let's IE8 problem be unresolved.

Will try to take time for this and#1505 to land tomorrow

@fhemberger when@gibson042 comments would be taken into the account, we're happy to merge it :-).

btw, thank you for all your patience with this!

Closed by828a718

Wow, great. Thank you! 👍

This PR was reverted on1.12-stable so I'm removing the milestone. We'll still try to fix the issues that popped up in 3.0.0 and not revert this patch there so I'm keeping the milestone on#1505.

(the issue to resolve is#2941).