Re: [http-state] Ticket 6: host-only cookies

[Adam Barth <[e-mail]>]

On Tue, Feb 2, 2010 at 11:14 AM, Dave Kristol
<[e-mail]> wrote:
> Adam Barth wrote:
>> Do you mean that
>>
>> Set-Cookie: foo=bar; Domain=example.com
>>
>> and
>>
>> Set-Cookie: foo=bar; Domain=.example.com
>>
>> are treated differently by user agents?  That might be true according
>> to 2109 (I'd have to check), but I don't think that's how user agents
>> work in practice.  They appear to ignore the leading dot.  If you have
>> a test case that show a difference, I'd be very interested in seeing
>> it.
>
> That is what I meant, and I do not have a test case.  That is what 2109
> said, and I believe (though it's been many years now... ugh!) I adopted that
> from Netscape's spec.  Generally speaking, the idea was that, in the absence
> of Domain=, the cookie should be returned only to the server that sent it.

I believe this is the line of code in Firefox that causes the leading
dot in the domain attribute to be ignored.

http://mxr.mozilla.org/mozilla-central/source/netwerk/cookie/src/nsCookieService.cpp#2159

There's a similar line of code in Chrome in the GetCookieDomainKey
function in this file:

http://src.chromium.org/viewvc/chrome/trunk/src/net/base/cookie_monster.cc?view=markup

Here's the comment:

  // Get the normalized domain specified in cookie line.
  // Note: The RFC says we can reject a cookie if the domain
  // attribute does not start with a dot. IE/FF/Safari however, allow a cookie
  // of the form domain=my.domain.com, treating it the same as
  // domain=.my.domain.com -- for compatibility we do the same here.  Firefox
  // also treats domain=.....my.domain.com like domain=.my.domain.com, but
  // neither IE nor Safari do this, and we don't either.

Note that Firefox now acts like IE/Chrome/Safari in regards to
Domain=.....my.domain.com

Adam