773132 - Implement auto-revoking blob URIs

Status: RESOLVED INVALID

(Core :: DOM: Core & HTML, defect)

Description

[Masatoshi Kimura [:emk]]

Per latest spec, blob URIs are automatically revoked by default after providing a stable state.

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

I'd prefer not using HTML spec's stable state for anything.

Comment 2

[Masatoshi Kimura [:emk]]

Then what should be used?

Comment 3

[Olli Pettay [:smaug][[e-mail]]]

I filed https://www.w3.org/Bugs/Public/show_bug.cgi?id=17758.
Looks like there can be also other problems in the spec.

Comment 4

[Masatoshi Kimura [:emk]]

I'm testing auto-revoke blob URI patch and I found it broke img-blobURI-2.html.
https://mxr.mozilla.org/mozilla-central/source/layout/reftests/svg/as-image/img-blobURI-2.html?force=1
I think this is expected because the outer SVG-as-an-image will not even start loading before a stable state (or a microtask checkpoint, it doesn't matter for this problem). But is it acceptable to potentially break existing contents which use blob URIs inside SVG-as-an-image?

Comment 5

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Yeah, loading the data-url source will definitely happen asynchronously, which means that the blob-url will be revoked by the time it's supposed to be parsed.

I think that's ok. At least I think it's at least worth trying, defaulting to auto-revoking definitely has a risk that it'll break too much content, but we won't know until we try.

Comment 6

[Arun Ranganathan]

(In reply to Olli Pettay [:smaug] from comment #3)
> I filed https://www.w3.org/Bugs/Public/show_bug.cgi?id=17758.
> Looks like there can be also other problems in the spec.

Olli: I'm not sure your problems with stable state are solved with microtasks -- that's equally vague, if not more so.

Can you look at that bug again?  I left a comment for you regarding "await..." versus "present..."

Moreover, there are now *two bugs* that I think legitimately address the problem of autoRevoke:

1. https://www.w3.org/Bugs/Public/show_bug.cgi?id=16953#c8 (and https://www.w3.org/Bugs/Public/show_bug.cgi?id=16953#c11)

2. https://www.w3.org/Bugs/Public/show_bug.cgi?id=17765

I think the above two bugs are the correct bugs with the specification, and NOT the microtask bug.  Thoughts?

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

Script (event listener etc) can't know what all scripts will run before the next stable state.
Microtask doesn't really have that problem.

Comment 8

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

I agree, we should definitely use microtasks

Comment 9

[Arun Ranganathan]

(In reply to Jonas Sicking (:sicking) from comment #8)
> I agree, we should definitely use microtasks

But microtasks aren't a real concept!  There's only the "magic words" which are "Perform a microtask checkpoint."  That's what this is all about:

https://www.w3.org/Bugs/Public/show_bug.cgi?id=16790

We could:

1. Bolster up the existing stable state requirements.
2. Go back to https://www.w3.org/Bugs/Public/show_bug.cgi?id=16790 and do a better job of defining what a microtask is.

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

Microtask is the outermost script execution.
MutationObserver callbacks get called at the end of the microtask or end of task, whichever is first.

Comment 11

[Glenn Maynard]

Please don't fragment the discussion: https://www.w3.org/Bugs/Public/show_bug.cgi?id=16790

Comment 12

[Masatoshi Kimura [:emk]]

Also,
  var fileurl = URL.createObjectURL(file);
  audio.src = fileurl;
failed to load because the resource selection algorithm is awaiting a stable state before reading a src attribute.
http://dev.w3.org/html5/spec/single-page.html#concept-media-load-algorithm

Comment 13

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

that seems like a bigger problem!

Comment 14

[Glenn Maynard]

Masatoshi, please see https://www.w3.org/Bugs/Public/show_bug.cgi?id=17765.

Comment 15

[Arun Ranganathan]

Auto-revoking Blobs are now harnessed to "post-script cleanup steps."  

See: https://www.w3.org/Bugs/Public/show_bug.cgi?id=19554#c4

And in spec., see: 
http://www.whatwg.org/specs/web-apps/current-work/#calling-scripts

Comment 16

[Andrea Marchesini [:baku]]

Autorevoke is not a thing in the latest version of the spec. I'm right? We should close this bug.

Comment 17

[Anne (:annevk)]

Yup.