[Suggestion] Do not use attachEvent to determine interactive mode

[normanzb]

some libraries, such as prototype.js or sscompat.js (which from script sharp) will add ms compatible attachEvent() function to HTMLElement.prototype, which might cause module failed to load in certain situation.

my suggestion is to determine the interactive mode by accessing script element's readyState, just like what curl.js did in their loader.

alternatively, we can try to delete the attachEvent from HTMLElement.prototype, see if we can successfully deleted it, if it did, then it means it is either IE or the attachEvent() method was artificially added.

[zjhiphop]

I'm also have this problem.Thanks!

[normanzb]

sniff the HTMLElement.prototype.attachEvent.toString() also works i guess

[jrburke]

I was looking at this version of prototype but I do not see where it does the attachEvent adding. Do you have a link for it?

Also, I'm not familiar with sscompat.js, I would appreciate a link for it too. This is what I found via a quick google search:

http://code.google.com/p/extsharp/source/browse/trunk/ExtJS2Samples/Web/App_Scripts/sscompat.debug.js?r=79#82

but that is a pretty old script.

I want to get a gauge how serious of an issue this is. I cannot protect against every script doing something potentially ill-advised in the page, particularly older scripts. But the higher profile ones I am willing to consider.

[zjhiphop]

sscompat.js is a base lib which build from script-sharp code.
And then we can write c#  and build it to javascript.But unfortunately,sscompat extended
some dom's prototype,such as attachEvent.Please see this:

    var attachEventProxy = function(eventName, eventHandler) {
        eventHandler._mozillaEventHandler = function(e) {
            window.event = e;
            eventHandler();
            if (!e.avoidReturn) {
                return e.returnValue;
            }
        };
        this.addEventListener(eventName.slice(2), eventHandler._mozillaEventHandler, false);
    };

    var detachEventProxy = function(eventName, eventHandler) {
        if (eventHandler._mozillaEventHandler) {
            var mozillaEventHandler = eventHandler._mozillaEventHandler;
            delete eventHandler._mozillaEventHandler;

            this.removeEventListener(eventName.slice(2), mozillaEventHandler, false);
        }
    };

    w.attachEvent = attachEventProxy;
    w.detachEvent = detachEventProxy;
    w.HTMLDocument.prototype.attachEvent = attachEventProxy;
    w.HTMLDocument.prototype.detachEvent = detachEventProxy;
    w.HTMLElement.prototype.attachEvent = attachEventProxy;
    w.HTMLElement.prototype.detachEvent = detachEventProxy;

So is this issue just happended in ie? 
What about this method:

           if ( /*@cc_on !@*/0) {
                //Probably IE. If not it will throw an error, which will be
                //useful to know.
                node.detachEvent("onreadystatechange", req.onScriptLoad);
            } else {
                node.removeEventListener("load", req.onScriptLoad, false);
            }

Thanks very much!

[jrburke]

OK, so it looks like this is the link for the project:

http://projects.nikhilk.net/ScriptSharp

I sent a contact form to the author to see if it is possible to use separate functions instead of modifying some global prototypes like so. While a fix here may help requirejs, it will likely not help other libs that do a similar type of capability detection. I'll wait to hear back from the author before considering a change. As I mentioned, I cannot protect against every conceivable change to the browser environment, and I would hope to work out a solution that works better for all libs.

[normanzb]

thank you for you feedback and investigation @jrburke :)

i checked prototype.js and it turns out i am wrong, they didn't add attachEvent to HTMLElement.prototype.

although prototype.js doesn't do this, i still feel that the code logic here is a little bit confusing.

the code logic implies that 'a browser who supports attachEvent must supports interactive mode also', is it always true? I don't see there is a strong relationship between attachEvent() and interactive mode, besides, compatible with script sharp can be a step to bring more c# developer to javascript world.

[jrburke]

There is no direct test for interactive mode that I know of, short of an expensive runtime test for it. I would rather use this sort of inference based on attachEvent than say browser user agenet detection though.

That said, I am hopeful that I can consolidate some of the logic for 1.1, but no guarantees. I'll put this ticket in the 1.1 bucket.

[jrburke]

Closing this in favor of using #193 since it has a pull request in it, and may go out if a 1.0.8 release is needed before 1.1.