W3C

VCWG Barcodes and Data Integrity

9 November 2026

Attendees

Present
antony_mott, dave_lehn, Dave Longley, elaine_wooton, greg_bernstein, ivan_herman, kevin_dean, manu_sporny, patrick_st-louis, ted_thibodeau_jr, wesley_smith
Regrets
-
Chair
-
Scribe
transcriber

Meeting minutes

Greg_Bernstein: Anthony, not too bad.

Antony_Mott: Hey, how are you?

Greg_Bernstein: Got back inance. Great. Great.

Antony_Mott: Really nice to see you on the call.

Wesley_Smith: Hey folks,…

Wesley_Smith: um I will give people a minute or two to trickle in and then we can go ahead and get started.

Meeting Overview And TAC Update

Wesley_Smith: I think we can go ahead and get started. We probably have a pretty full uh plate for today. Uh welcome to the November 9th meeting of the barcodes data integrity uh status list and forgery defense task force. Still on the market for a better task force name. if anybody has one. Uh so we did not meet last week and the week before that was TAC. For those of you who were there in person, it was a pleasure to see and or meet you. Um and we got some good stuff done. It was uh it was a productive week. We had some good discussion. Uh we had some great discussions in this task force's section of the VC working group meeting as well.

Wesley_Smith: Um, so I expect uh the first chunk of the call today I'd like to talk about TAC and what happened and if there are takeaways and directions and that sort of thing. Uh, and then we can go from there. Uh, does anybody else have anything they'd like to add to the agenda for today? Okay, that sounds good. Um, all right. So, we'll get into that stuff. Um, and then uh I I will not read that out loud every time. That would be even more difficult to say, but I appreciate it. And I appreciate that you you typed that all out.

TAC Resolutions And Horizontal Review

Wesley_Smith: That's that's impressive. Uh and I expect it was with your normal standard of of writing and grammar, so it's probably spelled perfectly. Okay. Um TAC went very well. Uh this task force we passed a couple of resolutions um largely process related to kicking off horizontal review for data integrity and a set of crypto suites. Oh, one another contender B IS Biffs. Yeah, SFIB. Um, this task force passed, well, the BC working group, uh, during the discussion of this task force's work passed a resolution to move data integrity in a handful of the simpler crypto suites to horizontal review, which is exciting. Uh, that is really great pro uh really great um progress and is only possible because of some of the awesome work that's been done over the last couple of months on those specifications.

Wesley_Smith: Um, the working group also passed a resolution to allow uh this group to more explicitly and concretely interface with other standards organizations and groups uh about the work that we're doing not just as individuals or a task force but on behalf of the working group as a whole. So that's really exciting. Um I I know that members of this task force have been doing a lot of work with other groups um both other groups within W3C as well as other at other standards organizations about what we're working on. Uh so it's good that we have now the official formal working group support behind us when we do that.

Wesley_Smith: Um on that note, uh I believe I will be presenting to some ISO groups on November 23rd, two weeks from today, um about a lot of our postquantum efforts that will involve things like the postquantum crypto suites uh as well as the forgery defense uh technology for postquantum or quantum resistance uh for documents issued before the quantum cryptocalypse.

Wesley_Smith: Uh, Greg Bernstein, go ahead.

Greg_Bernstein: What was the date again?

Greg_Bernstein: Because we're we we I can get to it if uh there was a couple updates. I just merged and and to get updated test vectors in, but there's some more stuff going on. When is when are you are they going to be looking at our stuff?

Wesley_Smith: Um, yeah. So, so I Sorry.

Greg_Bernstein: I wanted to make sure that our collective disclosure stuff is looking real good by the time they

Wesley_Smith: Yeah, that sounds awesome. Thanks. So, the date I requested is two weeks from today, uh, the 23rd. Um, I it will be a presentation to a technical audience, but I don't know a whole lot more than that. Uh, so potentially they could kind of drill down and focus on any number of different things. So, appreciate it. And you know speaking generally editors of any specifications that may be related to that. If you have in-flight work that could be relevant, try to get that nailed down.

Wesley_Smith: Um, Anthony, go ahead. Uh, no, no sound from you, Anthony. You may be quadruple muted.

Antony_Mott: Okay. Uh,…

Antony_Mott: let me try again. How about now?

Wesley_Smith: Yep. Yep,…

Antony_Mott: Yeah, somehow it works.

Wesley_Smith: I can hear you.

Antony_Mott: I just want to uh suggest uh back up what Greg was saying. We got a couple of changes. U of course Greg doesn't need me, but in case we needed it, I've updated the uh braking changes will be critical. Greg's probably talking about that, which was the ski sign ones. I know it's just one of the four algorithms, but I'll basically work with with Greg to try and help get that done, whatever needs to be done by the meeting you have with the ISO. And in particular, I think Greg, we just have to sort out a couple of the sentences that are changed normatively to see if the group has reviews or comments on that. Um, but it is within scope, I think, easily, right, Greg to get that done for the meeting.

Greg_Bernstein: Sounds good.

Wesley_Smith: Yeah, that sounds great.

Wesley_Smith: Uh, and just, you know, speaking generally about the timelines here, if if you have inflight things, especially related to things like ski sign, get them in if you can. Uh but also you don't need to uh you don't need to go too crazy on on any specific timeline. I think everybody working in the postquantum space understands that uh we're kind of you know building the plane as we fly it. Everything is very dynamic uh and changing very rapidly. So essentially every document that points at these things is sort of uh somewhat dynamic at least uh for the time being. So um that is that.

Wesley_Smith: If anybody here has anything else they'd like me to uh specifically go over with that ISO group, please let me know. Um, other than that, uh, did anybody else have anything they would like to talk about from TAC? Takeaways or anything along those lines? Manu, go ahead.

Manu_Sporny: Um yeah, that was a great uh summary. Thank you, Wes. And uh was wonderful to see uh a number of you in person. Uh wonderful meeting you in person for the first time. Uh Anthony, uh wonderful seeing Avon in person, of course, and Elaine uh as well as uh Wes. And uh sorry that uh some of the rest of you couldn't make it. Um uh I think the I mean things went fairly um things went smoothly in the working group. Um I think uh one of the things uh we wanted to make sure we did is get the horizontal reviews uh kicked off and moving forward for the core data model and the data integrity uh specs.

Manu_Sporny: basically data integrity, ECDSA and EDDDSA. What we're hoping is that the horizontal review for those will be very fast because not much has changed other than moving things around. there is the new um selective Merkel tree based selective disclosure mechanism Greg that that uh you've been working on um with some folks from Tencent um I think uh like that's the only you know big change um and we should talk about that but like so so I think the the goal is to try and get that into horizontal review and through horizontal review fairly quickly by basically telling them like hey not much has changed except us moving paragraphs around between the specifications.

Greg_Bernstein: Okay.

Manu_Sporny: Um, so my hope is that we can maybe I don't know I I think we agreed to do that at TAC. I don't know, Wes, if your read was the same.

Wesley_Smith: Yeah. Yeah. We we agreed we we agreed to bundle together base data integrity, ECDSA, EDDDSA, and the quantum resistant spec. So the quantum resistance spec being new uh will take the most review. Everything else we expect we can very explicitly point at a delta uh versus a previously reviewed version of the specification to hopefully get a quick turnaround.

Manu_Sporny: Yeah. And and we may want to split them into two different requests. one for the this will be a quick review and one for the take, you know, take some extra time looking at at this one. Um, I defer to,…

Wesley_Smith: Yeah.

Manu_Sporny: you know, I I my experience is that it might go faster if we do it that way versus putting it all together. Um, but, you know, I'm happy to have it done some other way if others feel more strongly about it. Um, the one thing I did want to talk about,…

Greg_Bernstein: finish. I think

Horizontal Review Process And Timelines

Manu_Sporny: I think one of the PRs that you raised, Anthony, around the Merkel tree stuff and how we're going to decide between uh, you know, shock uh, versus sick versus some of the Merkel uh, stuff um, for each crypto suite because um, I think that matters. But I'm I'm hoping we can talk about that uh today and then hopefully get to a point where like we as a group think like we're ready for horizontal review like in you know into and to go into candidate wreck. Um the other thing that we really need to start talking about is kind of test suites around this stuff.

Manu_Sporny: Um so anyway Wes I think this is largely just to like con what are the concrete timelines for getting barcodes for jewelry defense the data integrity specs and the quantum resistant crypto suites uh to horizontal review um because you know it's it's it's a gating factor that's

Wesley_Smith: Yeah, I'm just going to jump Q to say that barcodes already has been out for horizontal review and we've received review from everybody except the tag. Um, everything else does need to go out for review though. Avon

Ivan_Herman: Let's not forget about the forgery defense and the bitstring that's also under our core view is uh bit string is is obviously on the same bin because one single section has been added one or two properties if I remember well or one class and one property I don't remember uh that's all we have so we can make it very clear that you know otherwise we haven't touched a character uh forgery defense is new.

Ivan_Herman: So that's that's the one that has to go through the whole process.

Wesley_Smith: Yeah, I think it probably would be worth talking a bit about uh each each horizontal review request uh as a group right now.

Wesley_Smith: So, I plan to submit the forgery defense work for horizontal review as well as to respond to the review materials for the barcodes work. Uh, Kevin Dean, I know you are now a an editor of this status list. Would you be willing to submit the one version of that document for horizontal review?

Kevin_Dean: Uh not quite yet. There's a comment in um uh in my notifications that I need to take a look at. But um once that's I would say I would say yes,…

Kevin_Dean: I'll I'll take a look at it in the next uh uh in the next couple of hours and be ready in and…

Wesley_Smith: Oh, that is quite the turnaround time.

Kevin_Dean: be ready today.

Wesley_Smith: I didn't mean for you to drop everything. If you can get to it soon,…

Kevin_Dean: Sure. Um,…

Wesley_Smith: that's great. I mean more over the next week or two. Um, I know that, you know, there were a couple outstanding PRs. I'm not sure what the status of those are. That sort of thing. Uh, I mean, generally, uh, if you would be willing to take that out for horizontal review at some point over the next week or two, that would be great.

Kevin_Dean: guess I'll just need somebody to educate me on exactly what the process is, but you can message me separately about it. Okay.

Wesley_Smith: Yeah, absolutely. Yep. I'll send you an email that uh I believe Manu or someone shared at some point over the past six months about what the process looks like. It should be quite simple uh for Bitstring status list.

Wesley_Smith: It's a very involved process for new documents because you have to uh fill out some very substantial questionnaires and do a lot of analysis and that sort of thing. But because we already did this for well someone already did this for 1.0 uh ideally we can kind of short circuit all that point to the existing questionnaires and…

Kevin_Dean: Okay. Okay.

Wesley_Smith: the like and say all you need to do is look at sections you know 2.1 and 2.2 which are the additions that sort of thing. Okay. Thank you Kevin. I will get you that info today. Uh we also have data integrity and the crypto suites. Um Anthony or Greg, are either of you interested in submitting this horizontal review request? If not, I can certainly do it.

Greg_Bernstein: If you've got…

Greg_Bernstein: if you freshen mine and templates ready, that's wonderful. I'm assuming we'll um Anthony I Yeah.

Kevin_Dean: It's okay.

Antony_Mott: Yes, I sorry I should put my hand up.

Wesley_Smith: No, go ahead.

Antony_Mott: Um, yes, I uh uh that's one as well. Um I'd love to help if I can. Uh, Greg, if you wanted to do it, of course yourself. It looks like it's probably up to us to maybe to do that. But, um, I haven't done one before. So if I…

Antony_Mott: if someone could uh tell me what to do because it's different from the TR I did, which was interesting. So, I imagine there's a similar process for this. Um, but of course, Greg, if you want to do your cell phone, if you need a hand or if you want me to do it, I'm happy to do it.

Wesley_Smith: So that sounds great.

Wesley_Smith: Um, so what I'll do is, uh, Ke Kevin and Anthony, I will share you with you guys the instructions around how to do this. Greg, I can CC you, although you you don't need to do it. I I think that with this as a learning experience for you, Anthony, I think Montu's structure of separating out the two uh into things that have already been reviewed versus things that have not is a good idea. So, what we'll do is we'll start with having you submit uh the horizontal review request for the already reviewed bundle. Uh which is simple. And if that goes well, then maybe we can have you take a stab at putting a review request for the quant resistant stuff, which has not been yet uh reviewed.

Wesley_Smith: So that is a bit more involved. Sound good? All right. Uh Ivon, you have your hand up. Go ahead.

Ivan_Herman: Yeah, minor practical thing.

Ivan_Herman: Uh Manu has initiated this very good habit I think in the in the previous version of the burping group to have one place possibly one issue where the references to all requests are listed at one place. So one issue which says this is where we created the request for tag for privacy for whatever…

Ivan_Herman: because this kind of list is necessary later in an administration step when we move to CR or PR oh no not PR w PR doesn't exist anymore so please please don't forget to do that it it's immensely helpful later.

Wesley_Smith: Yeah. Yeah.

Wesley_Smith: Plus one. Uh let me go ahead as a quick example. And of course Anthony and Kevin, I'll share this with you. Here is an example for VC barcodes. Um it's an issue that is basically just a pointer to a bunch of other issues various repositories uh…

Ivan_Herman: Yeah, exactly.

Wesley_Smith: which which gives you kind of status of of those various things. Um okay that sounds good. So we have a plan for uh bing status list and for data integrity and its crypto suite children forgery defense uh and vcp is already done. So I think that is a that is every horizontal review assigned for this task force. Cool.

Wesley_Smith: Um, anything else folks wanted to mention about TAC or the action items coming immediately out of TAC?

Wesley_Smith: Anthony, go ahead.

Merkel Tree Subproof Advantage

Antony_Mott: Yeah, just um some good news.

Antony_Mott: Um I hope we might talk about this later. You might have had this on the agenda for later, but uh related to Greg and the Tencent contributor coming up with that. Um it's just such a huge thing. I hope Greg talks about it himself rather than me. that I was sort of the ambassador for the Merkel tree subproof. For those of you who don't realize it, it's uh it's almost an exponential improvement on the previous system we had. Uh I guess some of you realize that, but perhaps not everyone. And at the um conference, I talked particularly to the cryptographer. Mozilla had 31 people there or more. And there were some really smart people who said straight out, "Oh, it's a shame we can't do u the BBS equivalent."

Antony_Mott: And I was able to say,"Well, have a look at this." And, you know, I had a draft of that PR and pointed them to Greg and the 10-centent person's work and they were blown away. So, um, maybe Greg you could or if, um, where you could fit Greg if you want to. Uh, that was what just two or three weeks ago that you guys worked on that, but it is actually a gamecher. uh it's not perfect equivalent to BBS but um is essentially people don't realize how how useful and how efficient that'll make the the signings and things. So maybe um that's and it was another group too uh was yes the the the actual accessibility group members also didn't realize some of them are aware of BBS just didn't assume that it would be possible with postquantum. So anyway just that I didn't mean to blather on a bit but maybe Greg can fill in how much of an improvement that is and how that'll really help all specs moving forward into the postquantum world.

Wesley_Smith: Greg, go ahead…

Wesley_Smith: if you have something to say.

Greg_Bernstein: Uh I can give an up whenever it's appropriate on the agenda.

Wesley_Smith: Um, I don't think anybody else had their hand up for TAC topics, so go ahead.

Greg_Bernstein: Okay. So on on the so okay first of all put it into context. Everybody's been has to deal with these large signatures. So, so we did notice other people were doing kind of things like this.

Greg_Bernstein: Selective disclosure is a little special in the sense that we're not just doing we're revealing multiple items or not and you know in some kind cases we reveal the items right. So you know we haveation between these things and some of the existing ways do this um be quite efficient if you were only revealing like like one uh this was like the already standardized stuff around Merkel trees um which give theial reduction. Now the reason why we do this

Greg_Bernstein: is because we have way back we came up with two approach was signed individual statements. This is what we do which has very short signatures. We could do it for EDDDSA too. We set up this general rule. That's why when we go into cryptographic not when we go into horizontal reviews, you can tell I got something else on my mind right now organization.

Greg_Bernstein: when we go into the the moved almost all the algorithms into right so now all our other specs whether they're selective disclosure or not refer to the DI specs for the approach only ED CDSA SD will be different because some legacy things we were said hey why don't we just do this for everything and we that fights things between BBS and all of but ECDSA already disclosure. So one little change there. Now if you've got a very large signature that signual statements doesn't work because the signatures are so long.

Greg_Bernstein: So what you do is there's a general technique which hashes salts and hashes each statement and then you of all those statements and then you have to give them the person um the verifier this entire list of these which can get as many as you've got. Now the Merkel tree approach as it sounds uses a different technique and hey you know we're going to just send you we're going to sign the root of this hash tree and this is usually used for showing one thing was part of this structure and that can give you this huge exponential

Greg_Bernstein: advantage. Now with selective disclosure, we reveal more than one thing. So what we worked out there some people had looked at this um there um it's not like theoretical enough conference paper or something like that. But what do you do when you're revealing of these statements where if you do it stick point of view, you end up being worse than just the salted you do about half of the things.

Greg_Bernstein: However, you'd say, "Wait, this all looks like a sub tree." So, we drew out some pictures and things like that and said, "Oh, now there's nice doing." And the our friend from um said, "Hey, yeah, here." And I said, "Okay, let me try your nice recursive algorithm." And so, works out great. never be worse than achieved with the plain salted hash and usually much. So that's what that was about.

Greg_Bernstein: Okay, we used some constructions that were documented and some RFC's we applied other you know we looked at the problem we came up with some algorithms west Yes.

Wesley_Smith: Yeah. So the never be worse thing is is something I'm particularly interested in because it'd be nice to document that explicitly in the spec. Is that because the worst case, the base case is where you essentially give every leaf of the Merkel tree which is equivalent to the shock list, right? Yeah.

Greg_Bernstein: Exactly. Yeah. And that's where the uh this call like we called it it's like a subtree proof what we called it instead of an inclusion proof inclusion proof in these Merkel trees.

Greg_Bernstein: They tell get tree and what things to include and if you look at them and group them together and even if you try and do reduction it's not as efficient as looking at it as a subtree that's what we did and so that's why it's easy to prove you'll never be we yeah we got

Wesley_Smith: Ah, right. Okay. Yeah, understood. So, it's like for each Yeah.

Greg_Bernstein: the pictures and things like that. I can get you the the background. Wesley Smith:

Wesley_Smith: No, I I understand. Thanks. It's it's good to note um yeah if these things aren't written explicitly up in the text which I mean I'm sure they are. It's good to note things like this is never actually worse except for maybe it sounds like a a small constant factor but either right thank you

Greg_Bernstein: Yeah, I actually wanted to eliminate that constant factors. So, we have very deterministic tree structures.

Selective Disclosure Options And Complexity

Greg_Bernstein: I mean come from this nice simple deterministic algorithm that we just got from the that's why I was I mean it's nothing too very simplistic but it leads to tree now working more but there's a bigger issue you with that we should consider for the SD specs.

Greg_Bernstein: over at the um TF CFRG u we didn't we couldn't it was a little late to attach this stuff to the BBS core the blind BBS there's this concept that came up of closure and that allows you that allows the holder to say I'm going to reveal this value. Okay. Okay. And I send you the verifier what we call a cryptographic commitment and then I can prove something more about that. Like I'm to this value that's from my bank amount of dollars. I'm not going to send you that value. I'm going to commit to it. And then group that says I got more than $400.

Greg_Bernstein: So rule okay we approach lets us finalize the spec and things like that and other people are going to add on proofs without touching the spec in the selective disclosure case at with RDI specs. We use seabore encoding extra parameters. We just can't get around that in the um in the case of u digital statement case. We have things like public key. We've got like uh what's the other one?

Greg_Bernstein: the HMAC key. We have a bunch and that changes based on whether it's the sign individual claims shock or BBS with getting a lot of options and so it starts looking quite and we have to change the whole spec what would be nice and once again this is kind of the high level people uh maybe Dave

Greg_Bernstein: Manu work is if we can come up with a way to build in options with that seabore mechanism. So we and we're going to put in something that says we got more options and we don't have iterate the spec. Uh oh, my audio is dropping.

Manu_Sporny: It is ever so slightly.

Manu_Sporny: I can make out everything you're saying, Greg, but like it's like every tenth word drops.

Greg_Bernstein: Okay. Sorry.

Manu_Sporny: Um I don't know if it's a it's a you know it sounds like it's just dropping a packet every every couple of seconds. Uh on that um item though, Greg, the the options thing. I mean we can certainly do that, right? I we can we can fit options into the signature. Um and this is the thing that I wanted to comment on earlier. I think uh Anthony you raised a PR and I apologize I cannot remember which PR. I don't remember where I read it, but I think the PR that you raised allowed someone to select the selective disclosure mechanism in the signature itself. So, we had one crypto suite that would allow you to use like sick or shock and swip, you know, flip between them or something like that.

Manu_Sporny: Um, and I wanted to I wanted to So, one, I don't know if I misread that or not, Anthony, please let me know. But I if your PR did that, I think we should explicitly not do that. Uh, and the reasoning being um the more optionality we have in these signatures, and Greg, this speaks to the kind of the point you're making. the more optionality that we have in the signatures, the greater the chance that we won't have uh good interoperability for the same crypto suite and that will be used as a reason to not use data integrity, right?

Manu_Sporny: Um, so we need to I think in it just we need to if if we're going to put optionality like this in there and I I think it's it's already in there and I do think like optionality is a good thing to have in there um in as a general thing that can go in these signatures uh data integrity signatures meaning like whether it's ECDSA or EDSA or um uh you know u a postquantum or quantum resistant one uh that we have a generalized optionality mechanism in the signature but we should be very careful about the optionality that we provide.

Manu_Sporny: So for example um for MLDDSA uh allowing optionality between uh sick and shock is something we should not do right because well one I can't think of a good reason for why you would choose sick over shock for like MLDDSA um but even for other things uh like um uh like a ski sign I think we should pick a selective disclosure mechanism for the selective disclosure suite, right? We shouldn't just allow variability uh at that level. Curious to hear other people's thoughts. Um over to you, Anthony. Yes.

Greg_Bernstein: Jesus Christ.

Antony_Mott: my mic working?

Antony_Mott: Um, yes. Uh, yes. So, um, exactly. If you have a look at the PR and I understand it's pretty dense, but the only, um, the normative part that changes is exactly relevant to what you're saying. I actually thought the same that we might want to reduce it. But I'll try and summarize this. When we last met on the PR discussion a month or two ago, we didn't realize about Miracle Tree at that point that Tencent and Greg came up with. It's so good. Um, I probably should have made this clear in the PR, but I suggest that there's almost no situation where we would want to give optionality. So, agree with you, Manu.

Antony_Mott: In fact, I would suggest again this is up to Greg, Dave, whoever's more senior on that spec to just choose uh SH OC with Merkel tree. And the argument for that if you have a look at the table I adjusted, we used to have the various algorithms could have sic or SH, right? But in fact, unless I've missed something, the I'll call it shock with Merkel subree proof is so superior in every example. If we just make that the standard uh against people can weigh in with an array of one if someone wasn't doesn't want to do the batched issuance of it, we could actually literally have one choice. So does that make sense to anyone who's Greg? Also, might you weigh in on this? But I couldn't see a reason to have

Antony_Mott: the other versions with the postquantum signatures, they're completely unusable. And even just using ski sign and maybe Falcon, it's it's it's seems no point anymore. The Merkel tree um subproof is so efficient. So if that um summarized down to an array of one, what that means is the batched issuance. If someone doesn't want to put many copies as a reservoir and…

Antony_Mott: they're a little bit confused with that or don't feel they need to have unlinkability across presentations, they could choose a race.

Greg_Bernstein: Oh, got out too.

Greg_Bernstein: Not just me. Anthony, are you still there?

Manu_Sporny: Yeah, I lost.

Antony_Mott: Did I not did Did I not um ah okay.

Manu_Sporny: We We lost your audio 15 seconds ago, maybe.

Antony_Mott: Uh basically we could have one system or one crypto suite which has a batch size which could be as minimum as one or many if the reservoir size was large. Does anyone need that explaining or do you follow that's the way we try and get around uh linkability between presentations to have a reservoir of actual documents so that the um hashes are different each time. It's a great system. It's essentially suggesting that what Greg and the 10-centent person came up with is in fact the only way I'd recommend goes in the spec at all.

Wesley_Smith: Uh, Patrick.

Patrick_St-Louis: Uh yeah, I just wanted to chime in like I'm at this moment probably not well equipped to know and have an opinion on which one is the most effective or not. like this is something I'd trust uh that people working on the spec. But uh from my from the option of you know removing optionality and taking a feedback I can say I'm definitely in favor of this. I think it will make the uh I have a conformant implementation a lot more clear and evident into what that implementation supports and it could definitely speed up the process to you know building a roster of conformant implementation and it would make the testing a lot easier. So that's just my my two cents

Manu_Sporny: Yeah, I uh wanted to point people to what Wes said in the chat channel. Uh it is not as clear-cut as that Anthony and Patrick. You leak information with the Merkel approach, I believe, right? Like the total number of statements and if you're leaking that stuff, there's certain credentials where it's dangerous for you to leak that information. So for example uh if you have a credential and it lists the number of family members that you have um uh there is a clear benefit of sik over shock right uh sik does not list uh does not leak the full uh list uh of your you know full number of your family members whereas shock would so they're privacy concerns I think that we want to take into account it might not be

Manu_Sporny: enough uh for us to worry about that. Um but you know it is it is stuff like this that we need to take into account.

Greg_Bernstein: Oops.

Manu_Sporny: Um signature sizes also matter. Um right so like uh you know sik is clearly uh not as good if you want to use large you know a scheme that has large signature sizes and you want to disclose a lot of information. uh in fact it's you know Anthony to your point the shock approach is probably better in in even even for the smaller you know quantum resistance signature um uh schemes. Um the I think the other I thing to uh well actually there was another thing I wanted to make a point on but I can't remember it now unfortunately.

Manu_Sporny: Um yeah, it'll hopefully it'll come back to me. But but all that to say like the oh oh the complexity of the algorithm that's is one is clearly more complex than the other one. Um and so that's uh you know a potential uh concern uh as well. Um so I just want to make sure that I I I am guessing that we are almost certainly going to say like use shock for large signature sizes. Um, and then for the small like it I think it starts to become somewhat problematic when we get to much smaller uh signature sizes, ECDSA, DDSA, uh even uh ski sign. Uh we may want to have like a bit of a discussion um uh on that uh item as well. Um,

Manu_Sporny: Oh, the thing I forgot about Anthony was um issuing multiple things to be unlinkable has been suggested before. Um when it comes to deployments, it is not working out well. So it is not as again that item like of course we would want to issue you know a 100 different 10 different or 100 different um signatures for a certain credential and then you know the the application would select which signature to use or burn them. Um there are long-standing arguments against doing that. One of them is operations cost.

Manu_Sporny: all of a sudden you have 10x or 100x uh the signature uh times for MLDDSA and cost because they cost a lot of money a lot more than ED uh ECDSA um uh and ultimately it also requires the application to kind of keep track of which signatures have been used and which haven't. that has been an anti-attern in digital signatures for, you know, multiple decades now. And we'd have to argue against CFRG and and all the people that are like, you should never be doing that. Um, so so again, I I want to point out like it is not as clear-cut uh as it might seem at first. All of these things um uh come into play uh when we talk about the solution for each uh cryptographic suite uh we're we're discussing.

Greg_Bernstein: Excuse me. It's

Manu_Sporny: Um, that's a go ahead, Wes.

Privacy Considerations In Selective Disclosure

Wesley_Smith: Yeah, plus one to to most of what you said. A lot of it was what I was going to say basically that um I'm in favor of reducing the the optionality in a lot of cases here, but some of the language that we're using is pretty loose. Um and and it we need to be we need to be careful and clear about that uh especially in the specification. The the biggest notable thing being what I put in the chat that sik is better than the Merkel subree approach for privacy.

Wesley_Smith: uh the Merkel subree the amount of information it leaks sort of depends it seem uh without looking into it deeply it it intuitively seems like it does not necess in the worst case it is identical to shock and releases the exact information about your credential uh in the average case it puts you in a a bucketed binary search two to the n range um between like if you if you don't have to if the only info you have to give about half of your credential half of the tree is the top node, then you basically have given yourself an exponentialsized bucket for how many claims are in your credential. Sorry, that was messy. There's a range of how much information it leaks. Uh, in the worst case, it's the same as shock. In the best case, it is not as good as sick still, I think. But either way, we need to nail down exactly what that what that is because there is a meaningful difference. Um the I don't Anthony I think this is to your point

Wesley_Smith: about at least two I don't see a lot of value in uh shock at all once we've figured out what the if it is true that all of the constant factors introduced by the mark tree processing are essentially negligible um then both in terms of computational cost and in terms of uh information sharing and privacy uh the Merkel subree seems strictly better than shock but that that is that is what it is Um, uh, Manu, I think you said something similar to what I just said, but you said shock instead. You said we're going to want to use shock for cases where there are large Yeah.

Manu_Sporny: Yeah, I think I'm getting confused with the all the me things.

Wesley_Smith: Right. Understandably, I think you meant to say we're going to want to use the Merkel approach for large.

Manu_Sporny: Yes. Yeah, I thought I thought Merkel was a sub subset of the shock moniker. Clearly, I need to go and…

Manu_Sporny: reread the spec.

Greg_Bernstein: Yeah. Nope.

Wesley_Smith: I mean it it it is yeah it's just it's whether you put all the hashes in a list or…

Greg_Bernstein: Nope. There's not going to be separate.

Manu_Sporny: Oh, it is. Yeah.

Wesley_Smith: you put them in a tree basically. Um anyway uh oh pardon…

Greg_Bernstein: Okay. on on the information leakage. Go ahead.

Wesley_Smith: what did you say?

Greg_Bernstein: The information leakage issue.

Wesley_Smith: Uh I heard you say information leak le leakage. Is that a Was that a question at me?

Greg_Bernstein: I was going to actually uh clarify the information leakage.

Wesley_Smith: Oh, please do.

Greg_Bernstein: Okay. So in let's start with BBS.

Greg_Bernstein: BBS at the the cryptographic algorithm leaks the number of statements. End of story. Okay. Clinton tell people they can pad the number of me instead of call state terminology as messages. We tell them they can pad the number of statements. Current Merkel tree all for shock.

Greg_Bernstein: um or shock itself is based especially to be efficient back number of statements with the merry it's very easy to pad it out at the lower empty hash empty hash not against statements however the question where do you want a pad do you want to pad at the VC level player that's completely up to I want to do this but so you could say that the Merkel tree right now is equivalent to BBS and there's a question of where do you want to put in some padding if you want to put in padding

Wesley_Smith: Manu, go ahead.

Manu_Sporny: Yeah, that's a great point. Um, to follow up on it,…

Greg_Bernstein: Miss me. Yeah.

Manu_Sporny: I am I mean there's a calculation we're going to have to do, right? Um, and uh I would much rather do padding at the cryptographic meaning at the Merkel tree level than at the VC level. I think it needs to be just um uh uh uh you know uh opaque to people working at the VC level and then maybe as a part of the signature you know creation function Merkel tree creation function you get to say what minimum level of padding you want that feels like it then puts you into some kind of uh K anonymous you know set uh and and with that mitigation put in there then I think it is very

Manu_Sporny: clearly better to have the Merkel tree approach uh except when you are only revealing one attribute and I don't think that's a very realistic use case. Um like if you're doing that well anyway I don't think that's I don't think we should optimize for that use case. Clearly we need to hear from people that are implementing but it feels like we are saying like let's do the Merkel thing, let's pad the tree and let's do it at the cryptographic layer uh or at the Merkel tree layer and that seems like it would address all the concerns.

Wesley_Smith: Yeah. Just just one brief note, padding puts you in the bucket. I have between 2 to the n minus one and 2 the n claims in my credential which is great in concept but breaks down quite badly at very small values of n.

Wesley_Smith: Um so just something for us to keep in mind. Uh saying I have between three and four credential or claims in my credential isn't a very powerful form of privacy. Craig go ahead.

Greg_Bernstein: Oh, I was just going to say these things um we can BDS can learn from it as we can have this the Merkel tree is you just change the size of the tree and…

Greg_Bernstein: you fill in the extra pieces with um random values that would be your salts. We can offer the at the BBS side because once again BBS cryptographic says just give me messages and we can

Greg_Bernstein: and I then if we could be consistent between them how we choose to do that also already have uh we have that HMACization thing too. So, we've got a bunch of things that um and west answer your question though if we pat out the tree the very efficient so I mean if even if the if we pat out by double right the number you can't tell if it's a family of five or a family of 10 a family of two family of no kids right or anything you might want to say about those

Wesley_Smith: Um something sorry you cut out there uh for the last I had some trouble understanding

Greg_Bernstein: make sense.

Greg_Bernstein: Does that does that make sense? Um out the size it doesn't have to be a factor in size how you see how how these are constructed. They're not uh flat trees there.

Wesley_Smith: any you there you were cutting out for me Greg. Um so you're describing the considerations surrounding padding. What exactly are you noting about tree structure?

Greg_Bernstein: Uh we are using arbitrary binary trees not binary trees.

Wesley_Smith: Uh meaning that they don't have to be um exactly however many leaf nodes. They can be different heights, right?

Greg_Bernstein: So yeah,…

Greg_Bernstein: exactly. That's how we get the efficiency in the first place is we it's just a nice simple binary tree that's kind of like a a number with the number of levels.

Greg_Bernstein: It kind of corresponds to the decomposition and uh kind of corresponds to how the trees get structured. So it's a deterministic structure binary tree for arbitrary integer size and it works great. I mean that's why they chose it in that RFC that we

Wesley_Smith: Interesting. Okay. I I'll definitely need to take a look at some of the details here…

Greg_Bernstein: Yeah. Yeah. Okay.

Wesley_Smith: then because it sounds from what you just said, it sounds like my previous statement about the two to the n ranges is not true then because you can pad to sums of powers of two nodes instead of powers of two nodes. Um whatever I I'll take a look and there um I think we're all on the same page high level. We want to reduce optionality as we can. uh we know there is at least a tension between we we believe that in essentially all cases the Merkel tree way to do shock which I think is the correct nomenclature is better than the list of salted hashes way to do shock it's essentially strictly better both from a computational and…

Greg_Bernstein: Oh,

Wesley_Smith: privacy perspective uh however there is still there are meaningful use cases for which you would want to use individual signatures uh such as uh privacy Um, so we need to have ways to support at least two. Uh, and yeah, Mon, go ahead.

Manu_Sporny: Um and just to make sure I'm understanding this correctly, um there if if we use the the Merkel shock uh mechanism, then um it's just a single signature. So for MLDDSA, the signature size is just one MLDDSA signature. Compute is one ML.

Manu_Sporny: I'm seeing lots of thumbs up. Okay, that's great.

Wesley_Smith: shock. Shock is one signature that you're just signing different things.

Manu_Sporny: Right. Merkel.

Wesley_Smith: You're signing an array versus you're signing the the root of the Merkel tree or whatever.

Manu_Sporny: Yep. Yep. Yep. Yep. Um we would probably also want to come up with a comparison against um SD chot as well just to you know see uh see see what happens. Here's the reason I I think it I'm thinking of the uh supply chain use cases where you have bills of lighting that can have thousands of claims in it and it feels like this is very clearly uh better in that use case versus the standard just driver's license use case where you only have like 35 fields or whatever.

Manu_Sporny: But even for 35 fields, I'm wondering what the uh difference between SDJ and um and this approach is. Um this also raises the question on should we standardize sik at all. I think Greg, I mean, you know, it was refactored out and it's still used for the old ECDSA stuff and so because of that, that's a reason to keep sick around in the spec. Um, but we may want to it I think it would be wonderful if we could just say, you know what, for future selective disclosure stuff,…

Greg_Bernstein: We have two shock approaches.

Manu_Sporny: always use shock, the Merkel version of shock, unless there's a good reason not to. Um, and then maybe we do not standardize the list of salts uh approach because the Merkel one is just, you know, better in almost every better in every way. Yeah. Yeah. And I I guess we should not have two shock approaches if we can if we can figure out a way to do it.

Wesley_Smith: Did you mean to say we're not going to standardize sick? Manu plus one.

Manu_Sporny: Um I I suggested that the only reason to standardize sick is because ECDSA SD uses it and we should continue to use it. Um but for shock I'm saying we have two ways of doing shock today. Can we just use the Merkel tree one and drop the drop the other one?

Greg_Bernstein: Yes. Yes.

Wesley_Smith: Uh I will note that the the down I am sorry I join Q really quick.

Greg_Bernstein: Electric people don't want to do multiples. I swear.

Wesley_Smith: I am curious uh regarding standardizing versus not standardizing sick. that the issues with sick are um use case dependent. So is the fear that people will implementers will unintentionally you know it's a foot gun basically. The fear is that if we standardize sik people will use it incorrectly and that will lead to uh intractable whatever issuance and presentation times and things like that. Is that the concern? Because uh yeah that's my question anyway. They b me.

Dave Longley: I was just on the queue to to say I I felt like we had previously decided, but um in in this task force anyway to just have the Merkel tree flavor, but if we haven't, I think we should uh and it it sounds like we're all agreeing that we should only have the Merkel tree flavor of it. I think we should uh be really clear and say that going forward and drop the list approach. I don't think anyone has any good technical reasons to support

Greg_Bernstein: Keep going.

Wesley_Smith: All right.

Wesley_Smith: Uh, we are running out of time. Uh, I think we're all on the same page that we don't need shock list.

Greg_Bernstein: Peace. if I made a mistake,…

Wesley_Smith: We just need shock Merkel tree, whatever. Um, Manu, can you briefly respond to my question about sick? Um, I it seems to me that we still need sick, but I'd like to hear more about why you think we might be able to remove it. Um, Greg Bernstein, did you have a something to say about that?

Greg_Bernstein: but it had been how for a while. Um, I did merge the M because we had I had approved it a long time back and merged it in. So, right now we only have that one approach to

Wesley_Smith: All right, sounds good. Thank you, Greg Mercen.

Ivan_Herman: Yes.

Wesley_Smith: Um, Manu, go ahead.

Manu_Sporny: to your uh sick question. I think uh Greg, please correct me if I'm wrong here. We use sik for ECDSA in the new 11 one uh version maybe. Is that that's Yes.

Greg_Bernstein: Yes. Yes.

Manu_Sporny: Um okay. All right. So that that's the reason to keep it around,…

Greg_Bernstein: Very good.

Manu_Sporny: Wes. Yeah.

Greg_Bernstein: Works great.

Wesley_Smith: Yeah, I'm I'm arguing for keeping it around for…

Manu_Sporny: Um, okay.

Wesley_Smith: what it's worth.

Manu_Sporny: Yeah. Yeah.

Greg_Bernstein: It's It's got definite advantages.

Manu_Sporny: I'm not I'm not arguing strongly. Yeah. There's there's and we should not do this. So, caveat, let's not do this. But what we could do is basically say, you know what, we uh standardize ECDSAD, the old version of it. Uh, but the new version of it is going to use the the shockum approach and um it's not going to use the sick approach. I don't know what the the the bite benefits are. Um we could we could revenue ECDSA and EDDDSA crypto suite and use Shockum on both of them. Um but I don't I don't think we should take on that that new work. Um, that's

Wesley_Smith: Okay. Uh,…

Wesley_Smith: we're almost out of time, so let's keep it concise. Italy.

Antony_Mott: Yeah. Um,…

Antony_Mott: uh, it sounds like, um, from today I actually really need to also update the other two algorithms. We haven't talked about it. Greg probably knows what I'm talking about, but we really need to do those two other gigantic algorithms. I don't know if you guys are aware, but one of them is 27,000 bytes the signature. So, that would be a shocker candidate.

Antony_Mott: Uh, but just to get the go-ahad, I think you probably want that, right? So that the quantum crypto suite spec is complete with all four, not just the two we've done.

Wesley_Smith: Absolutely. Thank you, Anthony. The bong gohead.

Ivan_Herman: Yeah, I don't think that we should touch the ECDSASD by adding yet another possibility.

Ivan_Herman: I mean we already have too much uh we have a large number of crypto suit with all the various alternatives that would just add one more. We should leave it as it is in my view.

Wesley_Smith: All right. Uh, we are out of time for today. Thank you all for the discussion. That was very valuable. Um, I will talk to you folks soon and have a good rest of your week. Meeting ended after 00:58:38 👋 This editable transcript was computer generated and might contain errors. People can also change the text after it was created.

This transcription was generated by a large language model (LLM) and might contain errors. When in doubt, check the audio recording. This page was formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).