[webkit-gtk] Upgrading to newer version of webkit - need help.

Michael Catanzaro mcatanzaro at igalia.com
Fri Feb 6 06:32:41 PST 2015

On Mon, Jan 12, 2015 at 1:06 PM, Michael Catanzaro 
<mcatanzaro at igalia.com> wrote:
> I'm definitely interested in knowing which sites you've discovered 
> are broken with and without %LATEST_RECORD_VERSION in the priority 
> string, when -VERS-SSL3.0 is also present in the priority string. (I 
> don't really care what sites are broken by -VERS-SSL3.0, as it's not 
> reasonable to enable SSLv3 anymore and Firefox has already dropped it 
> as well.)

For what it's worth, it's come to my attention that by disabling SSLv3 
we have broken our insecure TLS protocol version fallback. That might 
be what's causing problems for you.

Normally when we connect to a server, we say "the highest protocol 
version we support is TLS 1.2" then the server says "too bad, I only 
support TLS 1.0" and makes a TLS 1.0 connection. 99% of servers on the 
web will look at the protocol version we send, then make the connection 
with the newest version supported by both client and server. But the 1% 
of servers that are TLS version intolerant see a version they don't 
recognize, have no clue what to do, and give up. When that happens, we 
as a client then try again, pretending to only support a lower protocol 
version. (This means the attacker can downgrade us to the lower version 
trivially by blocking the first connection, which is the only reason 
POODLE was so bad.) Firefox will try TLS 1.2, TLS 1.1, then TLS 1.0 
(and previously SSL 3.0). We used to try TLS 1.2 and then SSL 3.0, but 
now we actually tell GnuTLS to do a second connection with all protocol 
versions disabled, so it doesn't work. This should be fixed in 
glib-networking 2.44 (to be released next month, you can grab a 2.43 
development snapshot if this is important to you) which is smart enough 
to fall back to the oldest protocol version that we haven't disabled 
(currently TLS 1.0), thanks to work by Dan Winship. (I'm still having 
difficulty with one particular test site [1], but I think there is an 
unrelated problem.)

[1] https://bugzilla.redhat.com/show_bug.cgi?id=1177964
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-gtk/attachments/20150206/c227acf6/attachment.html>

More information about the webkit-gtk mailing list