Attach data objects directly to DOM elements

[dmethvin]

Only IE6 and IE7 are unable to garbage-collect JavaScript objects, I think the other browsers can handle cleanup themselves.

Switch to direct attach for those cases. This means we don't need to clean up data and events when elements are removed from the DOM, should be a perf boost.

This will require a change to global events since there will jQuery.cache will no longer contain all data for all objects and jQuery.event.trigger can't sneak a peek into the cache to trigger them.

Need a feature detect I can use to determine whether it's safe to attach data to a DOM object.

[dmethvin]

Deferred for discussion in 1.7; due to the remaining cases that need to be handled (i.e., removing special events and triggering global events) it may not offer enough benefits to be worthwhile.

[john]

Let's discuss this in the roadmap meeting.

[john]

Nominating ticket for 1.7 discussion.

[Rick Waldron]

+1,

[jaubourg]

-1, I don't see any possible feature test, hence why I'll vote negatively

[Rick Waldron]

I might be way off here but if we know that UAs that support deleting expandos are safe, then could we assume jQuery.support.deleteExpando is safe way to determine which browser to use internal cache (false) or attach directly (true)?

[timmywil]

+0,

[dmethvin]

-1, Based on my previous exploration, I don't think this will have a payback until we eliminate global events and possibly until IE6/7 die.

[john]

+1, I think this is fine to push for - global events can be tracked/handled separately (easily, even). And we can use deleteExpando to feature test it. I'm for it.

[scottgonzalez]

+1, if we've got a good way to detect and we get a perf boost, why not?

[ajpiano]

+0, This was Dave's idea, and he seems to have done the exploration proving that it seems unviable - but if we're willing to make the resulting architecture changes required, then it seems workable

[john]

Closing until we remove global events.