Internet Engineering Task Force 99, 20 Jul 2017

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF99-DCRUP-20170720-1040

Description

DCRUP meeting session at IETF99
2017/07/20 1040

https://datatracker.ietf.org/meeting/99/proceedings/

A

B

In one minute,.

C

Stop eating the food and sit down.

D

Curt, can you take notes this thank.

C

You anybody want to be a jabber scrub, we have a note-taker, you get a free gift.

E

C

Actually, it's they light up in different patterns. You get it's a rave in your hand, man.

F

B

This is Dee crow. If you are not looking for Dee corrupt you're in the wrong room, Lou sheets are sick, oh yeah, blue sheets are circulating I believe we have a scribe of his Kurt I'm. Sorry, you know Curtis taking minutes and a job ascribe is yo. Thank you very much. Introduction.

C

Okay, yeah welcome to decrypt the first meeting. You all know Marie and some of you know me: I, come from the security side of things.

B

We there was a Charter tweek, you may have noticed. This was merely because I noticed we weren't chartered to actually deprecated Shawa and so I asked Alexia to start that I realize in retrospect. That was probably me lessly pedantic, but I like putting him to work.

B

Please note. The note well well is very likes to say you have the right to remain silent. Anything you say can be used by the ietf. You have the right to consult with your attorney about your obligations and other privileges about intellectual property. This is the formal version note that the RFC numbers have changed. A new version of BCP 79 is issued. That's the RFC, eight one, seven nine, mostly it's a boiler, plays the same note.

B

So 15 minutes John. You have the floor.

G

We seem to have one open technical technical question. Please in the original. The this is part of the Charter. That I, don't think we have changed, is to add new and add new signing techniques to D Kim and the first, and there have been two and perhaps three of them, one is to add hash keys for RSA, because we, as we all know there, are bugs in DNS provisioning.

G

That makes it foolishly difficult to publish a key that doesn't fit in a single 256, byte next spring and and RSA key is bigger than one K bits. Don't do that so right, so the thought is we could do what other crypto protocols do and publish a 256 bit hash and put the key itself in the signature. That's technically straightforward, as far as I can tell there is no disagreement about that and Scot says he's working on implementing it. I think Murray said he would look at it.

G

I know I'm looking at it for the girl, stuff, okay, the second thing is to add elliptic curve signatures again. I, don't think, there's any any disagreement about that. We seem to agree that we're doing what's called pure IDI DSA, which is ev2 five. Five, one nine without an extra level of hashing and the signature is that's okay and the contentious question is: do we also want to add a hashed or instead we want to add a hashed version of of V V DSA signatures? Now the hash in this case there's no operational advantages.

G

The hash is the same size as a signature, and the signature is small enough that, even if that, even if a future.

G

Electrical curve algorithm had keys twice as big or written four times as big. It would still fit in a single expert. So my personal preference is that since the hash key says a hashed elliptical curve key adds no operational advantage and since Indy Kim, since there is no signature, since there is no negotiation between the sender and the receiver, a receiver must implement every possible crypto scheme that any sender might use to. Interoperate I would rather minimize the number of schemes and just add two other than three.

H

So Mountain Thompson, so there is actually potentially a security benefit to hashing, and that is that when you hash, you end up putting the public key in the message that you're sending and as a consequence of that that public key is then covered by the signature, and there are various ways in which signatures are used. But one of the less obvious properties of signatures is that if you generate a signature over over something, then it is possible.

H

It's certainly in RSA and maybe in in some of the elliptic curve variants, to generate a key that will validate with that same signature. So you can generate a message, sign it and then I can claim to a sent. That message, whether or not that's interesting or an attack that actually makes sense in this context is, is something that will require, like a huge amount of analysis, but I'm, just pointing out that the potential here for us to gain some additional defense against people. Switching switching keys out for messages just operationally.

G

We've had this is operation that we've had we've had these signatures for a decade and I. Don't think anybody's mentioned this issue until this month, so I. So if it is, if it's a practical problem, it's a mr. has to be a pretty low level one. It.

H

Might be a pretty low level, one I'll also point out there being signing messages with 512 bit RSA signatures and no one's really made not sure for us about that either. So, that's not really and persuasive argument for me.

I

Phil Han, Baker, yeah, III I think I need to see a example of quite what's being alleged here in that. Yes, there are some games that you can play, but that's kind of the right.

I

The reason that Dorothy Deming pointed out that we needed to have the hash in the signature in the first place, but that's what so I was a big proponent of hashes, because you can make the configuration experience easier for the user, because you know hash is you can make them small enough to be, but then again when I started to look into what you would want to manage a DNS zone or whatever, basically I, don't think that it actually does do any good, because what you probably want to have is some tool that is organizing your hashes and your keys anyway, and so it's not very, you should not be having to cut and paste the keys into the DNS zone file anyway, so that advantage went away so III I support.

I

Just you doing the Rakhi yeah.

G

Just just for people who aren't familiar with the am the unhappy czar not going away anytime soon, just for backwards compatibility now,.

J

So I'm channeling Jim Fenton saying if we are trying to minimize the number of things that need to interoperate do we need both hashing and elliptic curve. Aren't they two ways of solving the same problem.

K

So um just two I take what I think Martin's argument was on Eric rajala on about the potential ability which I also don't know how to exploit. Is that say um say we had some situation where a message was referred to um where, where you have an existing published message and what I want to do is I want to claim ownership of that message and I can't change the message by Ken: I can't publish mething stuff in DNS.

K

So what you do is you basically make a new key which generates the signature on the message, um and then you put that in DNS and how you can ask them it's one of those to a corresponding to this to say: I, don't know how to yet. But that would be the thing warrants concerned about. um We.

G

You would have to put put it in the same zone. Is the real key.

C

Ship of the message right, because it's it's not like an S mine signature on the content, you're.

K

Cleaning ownership of the the thing that music, this is exactly the same content. This.

L

K

Is a same issue?

K

Error found that acne right, it's the same, underlying underlying issue, um so right, um I, I, certainly um I, guess it seems to me that, like the natural thing to do, if you want to minimize the number of things you have to do- and you think you know we have to hash RSA- is to say let us let us go forward hashing all the time and then um and then I appreciate that, like 1024 bit, RSA has a nonce way is its way out, but um it was anyone to be it to finish.

K

But you know at some point arm which may not be too far in the future. um We may at the point where people simpler that world, where the large volume messages receive have two signatures on them, one of which is either eec or hashed RSA, at which point you can say, I'm tired of supporting 24, 10, 24 hours, RSA and listener, and because you don't have to accept anything in particular and because that what happens in receptors, you ignore it.

K

um Then this isn't like this isn't, like you know, receiving h.264 where you're like don't have internet video. If you like decide on what they uses before, is it matter? If you don't get like the beacon validity, that's what so it's perfectly easy to exam in the market and say well we're now at 78% of my messages or one of these two modern things and I, just don't care about twenty fours anymore. So um so I have to process the long tail.

K

So I think what I would have a key for is simply saying we are consistently hashing and then at some point, not lawsuits in future um until I say always in that far away, then you can just throw away them a hash code path. um So if we're finding them as code paths, that's axe. Isn't it because I say we have to me. If we think anybody's gonna do hashed RSA, then we have to have passion and you're talking when you say, but using the term hashing. You mean like.

B

Hash, the content inside and.

K

B

K

If the issue is about whether the keys ought to be hashed or not, oh, that's right, then then, then, then, consistently hashing is the thing that will lead you in the future to have only one go path.

H

Yeah not until some weather or whether or not the signing over the key is a value or not. I think the we have to consider the fact that the future is much bigger than the past and, as I could points out, we have. We have evidence to suggest that this is something that we don't have to worry about. A long tail, I agree with eken.

B

Anyone else others, sorry yeah.

G

I think I've talked to everybody who is doing an implementation. I, don't think any of us see see it's a meaningful simplification in in doing two hashes.

H

Is that fair, so I'm, not gonna like lie down on the road on this one I think if, if people are implementing this want to avoid the extra step, then then that's that's fine. A little little concern about the signing of the key I think this. That's part of the reason why we signed over identities in in TLS and various other protocols, but again, like I, said, couldn't, couldn't, pin anything to it. I.

K

I guess I will say that upon consideration, I, don't think it's worth having two forms of the EDD essay, so I think why I would prefer to hash it upon um and be clear I guess there to be two here pass: the key all the time. Thank you, the public key all the time. If people don't think, we see that then I think the answer is to do hashed, RSA and unhatched, so my rank order will be hashed, hashed key, always hash hash key for RSA on hash, key free, DVD essay.

K

um Finally, all finally hashing hashing and on hashing.

G

K

G

Hetero sex I think I almost agree with you, except for the universe.

K

G

For perhaps the order of some of the preferences I also.

K

G

K

Better right so I think I, guess I'm willing to defer I, think I and my people were thinking that sense of the room but I'm going to refer to the sense of the room like like.

H

The only other thing to consider is that if we have to go post column on this one, then we're gonna have large keys how's it gonna be hashed Oh, most likely I mean this is a rather speculative, but you know what one of the one of the techniques that the post called Oh months- users rather watch, yeah they're, all actually be large.

C

If we have to go post quantum, most likely TLS is gonna, be using.

G

The way on that, anyway, right yeah and if we startin before you go I'm just one comment like you're absolutely right, but if we have to go postpone and we're gonna have to revise this document anyway. So everything's.

C

Up in the air anyway,.

G

C

Could you post a note to the list explaining explaining an issue about the key swapping and then we'll just incorporate that into the suit direct trauma? Put that message you know in the security considerations. We're not doing this, we're not sure the risks, but here's what it is. Okay,.

K

As far as I can tell, what we are primarily arcing about is the is speculative future future simplicity of the code paths depending on. We think what we think is gonna happen in the future and if we think in the future we're gonna end up having to have hash keys in its. You have always hash keys, I'm thinking the future, we're gonna having largely unhatched keys that it's you start always on hash keys.

K

So that's to say, I think we all heard guesses on that willing to defer to the consensus the room, I, guess: I, don't know how you wanna I've ever heard of any people talk about it. So I think you know you want to go home or do you want to like whatever I.

B

Don't think we need to hum now I'm peppadews, let it go on the list and figure it out that way. If.

I

We go post quantum, we're probably going to be looking at a new version of SMTP, let alone TLS or whatever yeah yeah. If you look at what it would take, it's basically everything goes out the window so just a month or about bring.

C

I

F

H

We're done with this one, yes,.

C

C

G

Say it's a very clever thread is what I had when I hadn't thought up until you until you mentioned it right.

H

I'll provide text and before you disappear, John because oh I was just about to ask you: are you done the I? The review that I wrote of the draft hit the list I may have mentioned that I preferred hashing in it and that caused an enormous. It wasn't just you rat hole, yes, I, don't know who started the particular thread through your Apollo.

H

There are a bunch of other comments in there. That I think are really important. Yeah.

G

Although tonight it seems likely that we will, we will break break the draft up into separate pieces. So, like the comic the stuff about deprecating sha-1, it's an out no is now wasn't. It did.

H

Or didn't work for Belarus that that would be fine, oh yeah. The more verbose thing is a problem that I'll get to later, but the there are. There were some comments in there. That I think are pretty pretty crucial and I want to make sure that those don't get lost, particularly when we get things moving between documents, yeah I'm,.

G

Gonna close the loop on that yeah I missed the cutoff. Now that doesn't open again I will I will spirit its through them into the next version. Okay, great thanks. So.

J

G

J

Kind of a strange thing to channel, but well citizen channeling anything and did anyone find the stuff on this roof remark? If so, please contact.

K

So um I also just since I'm just a list a few minutes ago, but the names of these things, um so um is it as well yeah, both the DNS and the okay good. So if anybody I I, think anybody V Malia Josh the suggestion that the name should say to 55.9, where other than her 256 um and not just better than they should discuss it. But if not people remind us.

C

B

We are, we have extra time too, so we're little heads good, just as we start at 10 minutes early, so you can ignore it to some degree. The time limits on these all right, so the dekum usage document is supposedly and probably the simplest. These. The changes as I recall, are that we are deprecating use of sha-1 entirely. We are increasing the floor as a part and put it for the number of bits and keys that we should be accepting to 1024 and there's a third thing that I escapes me at the moment.

B

But essentially this is a an update to deke him directly that doesn't add anything but simply tightens up base requirements. Sorry, so Mike should I talk about that one. So yes, that's all this is Seth. Who just mentioned something is going to be shepherding. This document I think we're pretty close modulo. What we're about to talk about too less calling this it's it's a very simple update document, so glad Martin. So.

H

Martin Thompson, though there are a number of things that I had concerns with this document, I mean what you have on the slide is exactly it I think we should not do PSS in this document. Absolutely okay, that's so that's a no-brainer! Then we have two things and what you have right. There is almost the sum total of the document, yet it somehow manages to be five pages and two of those are in our fault but yeah.

H

Well, yes, but I think you might be able to get it down to two and I think that simply means providing some rationale for why the document exists and then saying must not use char. One must use keys of 1024 bits or higher, or you know, maybe we can make it a little higher than that even I. Think 1024 is perfectly reasonable at this point in time. The document currently includes a whole chunk of text from the DK MRC verbatim. Without modifying it, it's very confusing. We need to stop doing that.

H

Yep and the other thing is. We don't need to modify the a B and F, which is say have to tell people not to use those code points and modifying the ABN. F makes this document bigger and more complicated, and we have iona considerations and all sorts of other things that really don't advance things. We just need to tell people just stop using that.

B

It's broke I think there was a idea put in someone's head that the way you deprecated something is to pretend like it never existed before. Yes,.

L

B

We'll fix it.

B

But I'd, like I, said I plan to get this last called in short order. There's really not much to it. There's no reason yep anything else. So this one.

K

I'm just trying to think about what the security properties of all this are, so the existing, so I'm just gonna use for a second of people. Indulge me um so um these visiting attacks on sha-1 or all collision attacks, and we don't have any free image attacks. Yet maybe we never um so the the collision attacks on the so engine egg about like how do you support a collision in this system?

K

I think the way you exploit the collision in this system is and or no you say like why it can't because, like the produce, it is all produced by via pop by the center. But if the way you've put occlusion in the system is that um say you have like a mere relay that just spin filtering and now what I do is I. Give you two messages, one of which and I.

K

Can you throw any of the input right, but I give you two messages, one of which one for you to sign that is innocuous, and is it? Is it full my pornography or whatever, and then I take this inter off and I shovel to my party traffic, one on so I? Don't know like it anybody's written down what you might actually be ideal to do with a collision attack on deacon, but I'm I birthed someone thinking about that. I'm I don't know the system wanted that right now on the on.

K

The irrelevant point is that there are two more clocks: there's the clock when you stop making um someone's integers, which is the one where collision attacks, don't work longer become an issue and there's the point where you start perceiving them, which is the point where pretty much could become an issue and so and so on.

K

It might be nice in these documents to say something about what we think the times appropriate time scales were phasing out acceptance of sha-1, which I understand that we can't do for quite some time, but um I know I mean that the browser's made a lot of noise. Remember I stopped accepting sha-1 certificate, so I, wonder I, wonder if we should I wonder if we should say something at least about these. You know we expected to some point in future with some big time scale ways, but they actually say you stopped accepting them as well.

K

I think you just wrote our security considerations like um if somebody someone can actually so so as Adi refused to actually take any tokens, but it someone remind me to do it. I wobble okay. Thank you.

D

B

Be short, like the rest of the document.

G

On living, apropos of the Ayane stuff in my draft I move sha-1 to historic that we think that's a good idea. Yes,.

B

G

Would be one I, nted change.

B

G

I thought this did that, maybe that's not Thursday I.

B

Was that was in my draft.

G

I, don't think it's at Scott's graph. It should be in Scott, Scott, okay,.

B

Yeah thanks for that reminder. It's in my graft, I'm gonna, take it out of yours and yeah.

H

So they both drops include a similar sort of modification. If there is a way for I enter the flag, something is, you know, do not use. We.

B

Have a call exploit that we have a column in this in the register, we're talking about it, says obsolete or current. Yes, do that.

M

This is very I know some of the security related, some of the crypto related registries have things like should and should, plus, and something like that or whatever that says that the status is likely to change to be stronger in the next version. Let's not go, there says Martin okay, nevermind, like if it said. Thank you yeah.

G

John the meeting at mean, as far as when do stop except accepting them that's sort of more of a mug issue, is its most. Let me for ten years we told people sha-1 is bad and it's most in its appears to be largely a matter of figuring out who's still signing with sha-1 and beating up on the disiike. We can persuade them to change, which is not really the I ATF's remit.

G

You know and I think I, don't know, I, don't see how we can say anything other than the obvious like stop accepting them when there's when they're sufficiently. If you useful useful signatures.

N

Chris Newman a concern I have about the document is I'd like to see actually must check. The key size of the key is at least twenty four now I, don't care how you use the result of that that can be up to policy, but you must check this. The key size we had this I noticed we had this problem in TLS implementations, where things thought they were generating keys of the right size and they weren't and and it created Interop problem. So I'd really like to get that check in early.

L

Kurt Anderson responding to John's comment about people accepting sha-1. Yes, just the very fact that this draft is in flight, provided a lot of leverage and three of the major folks that were still actually sending sha-1 were convinced last month to stop doing it and have stopped doing it in the month. Subject. So.

B

Go true so going what has stopped.

L

And the other two are moving in that direction so going.

B

Back to D Marc, we don't need an experimental draft anymore right, yeah.

H

Sometimes I think perhaps I have a different spin on what what I was suggesting. Maybe this document should simply say must not, and let's not.

L

H

Or accept, and then it is a matter of policy for those who are deploying this and we can have in there security considerations text about when those things might actually start to need to be enforced. Is.

C

It, okay, if we say, must not generate and then just let it wither away. That's.

H

The challenge there is what Chris was talking about, which is that people end up generating 512 bit keys and.

L

H

And then you end up sort of in this situation. If people I think it's a useful hammer for people to have in the sort of policy and deployment area, just simply say the ITF ITF said, and so we have a way people can use that in their communities to actually have the things happen.

C

All right before we get to Phil is there anyone in the room has an objection to saying don't generate. Don't part, don't accept.

C

Okay, good segue, Phil, yeah.

I

I think that we should say it must not generate on the acceptance side depends upon what you're going to be using foot them for, if you're, using the signature as evidence of the message validity, as was used in the DNC attack, then of course yes, you're. Looking at the semantics of the message, and it really does matter if, on the other hand, all you're doing is doing spam filtering, which was the original idea of deacon, then you know breaking a sha-1 hash, breaking an RSA key that that looks more like a proof-of-work.

I

You know if somebody's going to break a sha-1 hash to to get a message through my spam filter. You know, maybe it's what so I think in general in terms of how the ITF works. If we go from must to must not on the receive side.

I

I think it sends a bad signal to ourselves in that it's tells us that we can tell the the community to turn on a dime and one of the things I've been weighted with sha Wan was that I was telling people to do sha to like a year after it was published by nist and people ignored me and then suddenly we have kab forum, say: okay, we've got to get rid of char wanna. We got to do it in 18 months and so I.

I

Don't like the idea that that's an acceptable thing to tell the community so I think that going through the phase of must not create, may accept- and you know the discipline of realizing that we can't just yank algorithms of out is useful because it will cause us to maybe be more careful.

B

In the future, so that it's interesting you, you you're like an interesting counterpoint to what we've been hearing, which is the community asking us to give them a hammer by which they can hit themselves over the head to move for it. Look there there's a there's, there's a certain amount of inertia that a must not accept will solve and so you're. Actually, this is the first question to give, which is it's not to say it's not valid but say they've been saying.

B

Could you please say must not, could you please say, must not I finally relent like okay? Well, if they must not, and now it's sort of the opposite perspective, it's interesting to hear both ways so Seth blank Val.

O

Mill, basically, it feels like we said, should not in 2010 before and there's been plenty of time for realize it to do that and the must 2007 so a decade and the must not language is what's necessary to hammer everyone else. It worked at MOG. Last month we had. You know there are three major ESPs who I won't name in shame, but who were still signing with sha-1. We talked to them and they all said with the threat of D Krupp and must not were just deprecating sha-1 language saying this is coming.

O

These conversations happening right. Okay, all three. We.

B

Go, we don't need to keep going.

H

On about must not too much longer about go ahead. Mom, Thompson, I think in response to these comments. The the key thing here is that the signature still exists and it can still be input into things like spam, filtering and whatnot, and if we have the right text in the security consideration is talking about. When this matters having the headline is still valuable, I mean I. Think people are asking for the headline headline thing is going to be. The body of the document must not blah blah blah and then there's gonna be some deployment considerations.

H

That say you know this is gonna, take some time and people are going to still be signing with them and there's still value in having that signature there in some contexts, but I think it opening to to back away fully from this thing: yeah not already, there's.

C

Also, a timetable from publication to when it actually gets approved.

H

And then then the card has to go.

P

Coonan we had this discussion in other working groups also operating Saarinen, and in most of the case, you have to understand. There's two different things we are talking about here: there's must not implement and must not use. Those are very different things implemented or actually the vendors who make those products they hate must not implement. People task means that they immediately. If they want to be conforming it this documentation, they have to remove the whole code, which means that then they can't have a backward, compatible, plat flag. I said oh yeah, but yeah.

P

We have to see this product here that doesn't support it, so we have to have it and they say sorry, we can't do if it must not implement. Then you can have. You know like, for example, OpenSSH has this, you know if you want user group 1 and solve one, you have to enable these flags to be able to use it, but you can still use it must of use. Is okay, I think, that's that that's something that we can say.

P

But of course, usually we don't give instructions, but you can use because we say that it's actually, you know policy policy requirements for them. You know these people who are using it must not implement is really annoying in that sense. That, and if you say oh, we can burst not with some kind of oh, you can use it. If you want it, it should not. It's not much, not anymore.

P

If you have any reason why you want to use it, then it should not with some reason- and it's always good thing to say the reason why you wanted, but then you can use it if it is not well.

C

Cut the line after Kurt yeah yeah.

Q

I only want to about must not under should not have the experience that all these should or should not in the ITF drafts. Also only our justification to to not move forward.

Q

I'm more often wish that there is a must not also in that case and everybody who can't follow that must not is free to not use months. Yeah.

L

Kurt Anderson, so, given that we all realize that there's a certain cadence to getting documents out and published and and updated, maybe we can write a timetable for moving through these phases into the document. So, as was suggested, we start off with a must, not send, then we add. Maybe we put a deadline next year. June of 2018 must not accept and we write that into the document so that there's essentially a tripwire in there and Ayana changes the registration in June of next year. That says, must not accept in the meantime.

L

It so should not, and it becomes a must not, and everybody can point to that they both get the headline as well as we don't have to do another document and it just happens so.

C

What I've been hearing in the room so far is most people are in favor of just making, it must not generate, must not accept all right. Let's take a home. How many the question is, how many people are in favor over the documents saying, must not generate or accept and how many people are not in favor and how many people don't have enough info to decide. So, all in favor all those opposed all those who need more information that seem pretty strong and favor yeah they're great yeah.

J

C

Standard ski will confirm it on the mailing list, but.

C

Don't we have to confirm I'm a list? That's when we do it insecurity. That's not the point.

M

The point is, it sounds like you just took a vote.

M

C

M

Have their objections been aired and discussed, or did.

L

M

J

Okay, so Fintan and jabber says that he agrees with must not generate, but it's too soon for must not accept.

J

R

Well, you, you got the sense of the room, so you need to take it to the to the mailing list of make the final decision and let people the chance to object one where the other. So it's okay to say there was a room consensus. You know rough consensus, but you know it's no more than that. Mm-Hm.

B

Okay, so apart from me, not must not peace of this conversation, I think we've covered all the known issues. With this draft I sent her some. My own review comments. I'll get Scott to do a new version.

B

No he'll be able to hear that and see the notes of this or hear the transcript and figure out what else we discussed, but I I do plan to get this sort of last call this fast as I can reasonably get it done without missing any important points. There's really not much to this left, the ECC I, don't think Scott is here. Is he Scott Rose, oh yeah, I'm? Sorry,.

S

Hi I'm Scott, Rose, NIST, basically, I. Don't think this draft is needed anymore Amy, because this is for there's a community out there. That needs to also implement FIPS 140 stuff. Having some internal discussions, it seems now that there's actually a roadmap for this sort of thing to get EB DSA into FIPS 140, so things are kind of progressing barring any kind of strange management stuff. So, in effect, the actual reason this draft exists is gone.

S

If there is a bunch of people that really really really like ECDSA over the kerbs mentioned on the list and otherwise I'm not gonna work on this anymore, we.

B

Can label it dead, yeah.

S

D

B

Wow, no, you got another four minutes.

B

Sorry I'm, really! Sorry I'm.

T

Very, very, do you want me to say it again.

C

Wow, that's that's actually seriously great news that too Feb 19 and will show up in 5th cool.

B

No, no, no you're! On the hook now.

H

You have to include the standard disclaimer that you're not making announcement about future product uncle bub.

B

We are done unless anybody else has other business for this working group.

B

Okay, we may or may not see you in Singapore. Thank you, everybody. Where is it.

S

F

E