From: full-disclosure at royds.net (Bill Royds)
Subject: Re: Internet Explorer URL parsing vulnerabi lity

The real problem is the http: URL syntax.
Good design does not allow secondary attributes (account and password) to
precede  primary attributes (hostname) and the convention of URI query
arguments allowing anything after the ? Confuses it a bit more.
Consider these URL's
http://www.microsoft.com@....slashdot.org
http://www.microsoft.com?@....slashdot.org

The RFC says that the first is a username www.microsoft.com followed by a
host www.slashdot.org
The second is a hostname www.microsoft.com followed by query parameter
@www.slashdot.org 

That is extremely bad language design because similar syntax has entirely
different (and opposite) meaning.
The whole scheme needs to be deprecated and a better one invented.

What this whole thread is about is that browsers and people both
misinterpret it. Would it not be better to fix the syntax.
At least allow browsers to ban this username:password@...t syntax and allow
them to prompt for username:password and use a different HTTP command other
than GET to send it for authentication
At present most usage of http authentication is after a 404 error which then
sends information in the headers so it is not actually used much even in
cases where authentication is needed.
Perhaps a separate authenticated method AGET that would allow it instead of
being allowed on all HTTP methods would be a start.
Then any site getting an AGET instead of a GET would know that
authentication was coming.


-----Original Message-----
From: full-disclosure-admin@...ts.netsys.com
[mailto:full-disclosure-admin@...ts.netsys.com] On Behalf Of Nick FitzGerald
Sent: December 11, 2003 4:29 PM
To: full-disclosure@...ts.netsys.com
Subject: Re: [Full-Disclosure] Re: Internet Explorer URL parsing vulnerabi
lity

It was written (by whom doesn't really matter):

> Check that. With Moz 1.5:
> 
> Opening in a new *TAB* takes one to MS. Clicking the link takes one to /.
> with "http://www.microsoft.com%01@...shdot.org/" in the address bar.
> 
> That's odd.

Not at all.

Can you not read HTML source?

The page has an href anchor tag (to MS) and a script (with a %01-
obfuscated URL to /. that "implicates" MS) on the onclick event for the 
anchor tag.

Thus, clicking the link _IF YOU ARE SILLY ENOUGH TO HAVE SCRIPTING 
ENABLED_ activates the script that implements the "trick" URL.  
(Almost) anything else you do in Moz (or a Moz-derived browser) to 
access that URL will result in the script not being activated and the 
plain URL in the href argument of the anchor tag being "seen" and/or 
acted on instead (that is why MS' URL is seen in the status bar ("task 
bar"?) when you float the mouse over the URL).  You should now be able 
to work the rest out.

...

In general, there have been a lot of really badly misinformed comments 
in this thread.  Things that suggest the poster does not understand the 
userinfo part of the URI RFC; things that suggest the poster has no 
idea that the "left hand URL" is not a URL at all; and more.  Please 
folk, if you don't know how something works either _ask_ or sit back 
and read (as the odds are someone will explain it all in plainer 
language or the penny will otherwise drop within a few more posts 
anyway).  If you are not absolutely sure that you understand how it 
works, don't post "it works in mozilla" (when it clearly does not) or 
any of the other myriad (near) clueless responses we've seen.  Clueless 
posts add substantially to the nose and can greatly increase the 
workload of folk who are now worrying about what, if anything, they can 
do to reduce their exposure to this.

Cheers...


Regards,

Nick FitzGerald

_______________________________________________
Full-Disclosure - We believe in it.
Charter: http://lists.netsys.com/full-disclosure-charter.html

Powered by blists - more mailing lists