Add back XMLDocument.prototype.load in Gecko compatibility mode

[domenic]

Opening this to track the discussions in #1478 (comment) so I don't forget. We'll want to add some links to the Bugzilla bugs @bzbarsky referenced explaining what's going on here and stating something like "as of 2011, this was necessary; if anyone using a Gecko UA string has evidence that it is not, let us know."

We could still make XMLDocument === Document (see whatwg/dom#278) and just add load to Document, I believe.

[zcorpan]

It may not be web compatible to have load on Document. Don't do that ☺

[annevk]

Yeah, I don't think it is. That's why we added the distinction. Perhaps it might be in non-Gecko browsers? Ugh, this all sucks.

[foolip]

When I looked closer in whatwg/dom#308 (comment) it didn't look like problem with the Sarissa library from 2011 actually depended on XMLDocument.prototype.load. It could still be needed for some other reason, but how to find out?

[foolip]

I've tried to summarize everything I could find about document interfaces here:
https://gist.github.com/foolip/103963a1ae8598d2baedd296f4a1bf4c

Since the discussion is spread out, I arbitrarily suggest discussing the larger issue in whatwg/dom#221

[ExE-Boss]

Firefox has bug 332175 open to remove this, and no other engine implements this, so WebCompat breakage as a result of closing this will probably be virtually non‑existent.

[bzbarsky]

@ExE-Boss What makes you say that there won't be webcompat breakage, exactly? People were using this together with UA-sniffing, last I checked. And the bug you reference was opened 12 years ago, and explains all that. Do you have any actual data showing there won't be breakage?

[ExE-Boss]

https://telemetry.mozilla.org/new-pipeline/dist.html#!cumulative=0&end_date=2018-08-30&include_spill=0&keys=__none__!__none__!__none__&max_channel_version=beta%252F62&measure=USE_COUNTER2_DEPRECATED_UseOfDOM3LoadMethod_PAGE&min_channel_version=null&processType=*&product=Firefox&sanitize=1&sort_keys=submissions&start_date=2018-06-21&table=0&trim=1&use_submission_date=0 shows only 0.003% of all page loads using XMLDocument.load(), so I’d say it’s virtually non‑existent.

[annevk]

That's still a relatively high absolute number and depending on the breakage might not be acceptable. It's perhaps worth trying again, but that's not good enough data to suggest Firefox is going to remove this to standards forums.

[foolip]

Yep, 0.003% is a small number, but when we remove things from Chrome, that's still large enough that we want to understand what the breakage would look like. If it would throw an exception in a script that critical to the functioning of the page, for example, then that wouldn't be a negligible problem.

[zcorpan]

load and async have been removed in Firefox 68
https://bugzilla.mozilla.org/show_bug.cgi?id=332175
https://bugzilla.mozilla.org/show_bug.cgi?id=1328138

Both of these are tested in dom/historical.html

I think we can close this issue.

[bzbarsky]

load and async have been removed in Firefox 68

Note that this has not shipped and we have no idea whether this is web-compatible yet...