[webkit-reviews] review requested: [Bug 6272] XMLHttpRequest freezes on getting a missing document with overridden Content-Length : [Attachment 12426] proposed fix

bugzilla-request-daemon at macosforge.org bugzilla-request-daemon at macosforge.org
Sun Jan 14 07:29:06 PST 2007


Alexey Proskuryakov <ap at webkit.org> has asked  for review:
Bug 6272: XMLHttpRequest freezes on getting a missing document with overridden
Content-Length
http://bugs.webkit.org/show_bug.cgi?id=6272

Attachment 12426: proposed fix
http://bugs.webkit.org/attachment.cgi?id=12426&action=edit

------- Additional Comments from Alexey Proskuryakov <ap at webkit.org>
> Really should leave out those parameter names ("request", "error", "r").

  Done.

> I could see just using a HashSet<StringImpl*, CaseInsensitiveHash<
> StringImpl*> > instead.

  Changed (used String instead of StringImpl* though).

> Where did the list of forbidden headers come from? How are you sure you have
> the right list?

  They come from the XHR spec draft, plus Via from the Firefox implementation.
As mentioned in comment 9, we may want to discuss this list on our own.

> We could put something on the console there. Why not do that, instead of
having
> the FIXME?

  My thought was that we'd need to know the script line and URL to properly log
the warning. But this is a general problem elsewhere in WebCore, and logging
"something" is indeed better. Added.

> Why not add SECURITY_ERR with this patch? Is there a risk that if after add
it
> to DOM Core, the value will change?

  Yes, this was my concern. In Hixie's proposal, it's just has the next unused
code (18), so a confict seems possible.

> And one more comment: I think we may need a Dashboard quirk to not raise an
> exception in one or both of these cases, and instead fail silently.

  I very much hope that it won't be needed. Unlike previous quirks that deal
with coding mistakes, this would create a genuine difference in runtime error
handling. Please note that blocking dangerous headers is silent per the
standard anyway, so there's just the NETWORK_ERR case to worry about
(admittedly, a rather significant one).



More information about the webkit-reviews mailing list