Re: [TLS] I-D Action: draft-ietf-tls-negotiated-ff-dhe-10.txt

So this is the exact problem I've been talking about.

After this draft is actually implemented in all TLS servers everywhere,
server operators can add Java-specific errata to their configuration to
work around the fundamental problem that DHE is breaking Java TLS
handshakes.

Or you can just add the Java property I cited above. Which is total oddball
errata. Nobody should have to worry about jdk.tls.disabledAlgorithms="DHE
keySize > 2048"

This is just adding to the "headache" of maintaining a modern TLS stack.

I'll freely admit I just made mistakes trying to diagnose this problem but
as a practitioner I seriously don't care about supporting these things and
want them to just go away and for TLS handshakes to just work and not
spontaneously break because of misconfigurations.

I don't see how adding more complexity fixes the problem.

I just want DHE to diediedie.

Perhaps someone can explain how keeping DHE around is actually beneficial
in any way whatsover? Right now it's just giving me a headache and making
me angry.

On Tue, Jun 2, 2015 at 11:54 PM, Yuhong Bao <yuhongbao_386@hotmail.com>
wrote:

> The idea is that the server can send a different weaker DHE key in the
> ServerKeyExchange, or not use DHE suites at all.
>
> ________________________________
> > From: bascule@gmail.com
> > Date: Tue, 2 Jun 2015 23:52:36 -0700
> > Subject: Re: [TLS] I-D Action: draft-ietf-tls-negotiated-ff-dhe-10.txt
> > To: yuhongbao_386@hotmail.com
> > CC: geoffk@geoffk.org; tls@ietf.org
> >
> > Yes sorry I meant ClientKeyExchange...
> >
> > But can you explain to me how this solves the problem for legacy clients?
> >
> > On Tue, Jun 2, 2015 at 11:52 PM, Yuhong Bao
> > <yuhongbao_386@hotmail.com<mailto:yuhongbao_386@hotmail.com>> wrote:
> > The client don't receive the ServerKeyExchange message containing the
> > DHE key at all until after they sent the ClientHello.
> >
> > ________________________________
> >> From: bascule@gmail.com<mailto:bascule@gmail.com>
> >> Date: Tue, 2 Jun 2015 23:35:46 -0700
> >> To: geoffk@geoffk.org<mailto:geoffk@geoffk.org>
> >> CC: tls@ietf.org<mailto:tls@ietf.org>
> >> Subject: Re: [TLS] I-D Action: draft-ietf-tls-negotiated-ff-dhe-10.txt
> >>
> >> On Tue, Jun 2, 2015 at 11:32 PM, Geoffrey Keating
> >>
> > <geoffk@geoffk.org<mailto:geoffk@geoffk.org><mailto:geoffk@geoffk.org
> <mailto:geoffk@geoffk.org>>>
> > wrote:
> >> It's covered in section 4:
> >>
> >> If at least one FFDHE ciphersuite is present in the client
> >> ciphersuite list, and the Supported Groups extension is either absent
> >> from the ClientHello
> >>
> >> Unless I'm mistaken, unless you configure the
> >> jdk.tls.disabledAlgorithms property explicitly (with e.g. "DHE keySize
> >>> 2048"), Java clients are aborting *before* they send the ClientHello.
> >> Please let me know if you're seeing otherwise. I could be mistaken and
> >> perhaps there's a server-side workaround for this that isn't "disable
> >> all DHE ciphersuites". But this is what I've personally observed and
> >> have been advising people about.
> >>
> >> I'm not saying it can't be fixed with additional
> >> configuration/errata/etc, I'm arguing that it's *breaking clients in
> >> the field right now*
> >>
> >> tl;dr: I am seeing *widespread TLS breakages* because of this resulting
> >> in *huge outages* for Java clients
> >>
> >> --
> >> Tony Arcieri
> >>
> >> _______________________________________________ TLS mailing list
> >> TLS@ietf.org<mailto:TLS@ietf.org>
> https://www.ietf.org/mailman/listinfo/tls
> >
> >
> >
> >
> > --
> > Tony Arcieri
>

-- 
Tony Arcieri