304260 - Nothing documents that nsIClassInfo.flags == nsIClassInfo.DOM_OBJECT will trump implementing nsISecurityCheckedComponent

Status: RESOLVED WORKSFORME

(Core :: Security: CAPS, defect, P5)

Description

[timeless]

this is really messing me up, and I think it's probably wrong too.

Comment 1

[timeless]

Attachment: reorder classinfo after securitycheckedcomponent

Comment 2

[Boris Zbarsky [:bzbarsky]]

I'm not happy with the perf impact of QI to nsISecurityCheckedComponent for every security check on everything.  In particular, for Elements that will be VERY painful....

Comment 3

[timeless]

bz: well, what alternatives are you offering?

should something be responsible for caching that information?

this stuff causes real pain for js objects that need to act like dom objects but want to be able to have their own control over how security behaves (which doesn't seem unreasonable).

Comment 4

[Boris Zbarsky [:bzbarsky]]

If I had a counterproposal, I'd offer it.  All I can tell you is to expect a 3-5% hit on Tdhtml from this change if you land it as-is.

Perhaps the right thing is to revisit our use of classinfo for security checks?  Or revisit the use of nsISecurityCheckedComponent?  Or some other rationalization of our over-complicated security architecture?

Comment 5

[Boris Zbarsky [:bzbarsky]]

How about a classinfo flag to indicate that the class implements nsISecurityCheckedComponent?

Comment 6

[BugBot [:suhaib / :marco/ :calixte]]

The bug assignee didn't login in Bugzilla in the last 7 months.
:ckerschb, could you have a look please?
For more information, please visit auto_nag documentation.

Comment 7

[Boris Zbarsky [:bzbarsky]]

None of the code this bug is about exists anymore. In particular, there is no nsIClassInfo::DOM_OBJECT thing, there is no nsISecurityCheckedComponent, and there is no property-access checking in the script security manager.