Internet Engineering Task Force RATS Architecture Design Team, 16 Mar 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: RATS Architecture Design Team 2021-03-16

Description

RATS Architecture Design Team 2021-03-16

A

Morning, good afternoon, elliott, hey.

B

C

D

Okay, we just call you all thomas.

E

Collective name, it mean twins.

F

That would be funny honey, hey I'll I'll, send you my mortgage thomas.

D

Two I'll send you mine,.

F

No, you you, you send me your crypto currencies.

G

D

But I want you to send me: elon musk's, cryptocurrencies,.

D

Yeah I paid off my I finished paying my mortgage uh a year ago: oh yeah yeah. It's pretty amazing.

D

D

We have these, we have a uh well americans get to deduct their interest rates. I don't know what happens in europe. That's not the state case in canada, but canadians get to take their house as a uh tax. Free um capital um gain.

D

So, as a result, it turns out it's worth borrowing money to invest it to pay off your mortgage.

F

So like so, if you sell it in the future, there's no capital gains tax.

D

On your primary residence only.

F

D

F

uh What is it called like inheritance tax? If you hand it over to your kids.

D

uh They pay tax on the house on the estate yeah.

A

D

um I think, but but if you're smart- and you know that you're dying soon, then you can sell it to your kids um prior to your death at a very advantageous rate and then, of course, their house, and they they didn't. You didn't have to do anything um yeah. I think there's something like that. Anyway. um Welcome yogesh and penguin to our group.

C

Thank you. Thank you, everyone. This is a mobile corporation and I'm very interested in this rights working group and I'm here to see if I can do something to help or get a promote understanding of this subject. Thank you.

G

Hi, I'm yogesh and I work closely with thomas fossati, whom you were conversing few minutes back. So I also work in verizon, open source project and look forward to be part of this conversations.

D

Okay, wonderful, here's hank.

H

Sorry for that overlapping meetings due to time zone changes, so I had to stay a little bit long on the other one.

D

All right, thanks for joining us, since you have two of the three pulls: no you've only got one, but all these issues all right, so um we're gonna walk through uh for sure this one um and I think we're gonna have to spend some time on um understanding the issues for the other ones. um Unless dave, you think this one is such low hanging fruit that we should just deal with it. Now.

D

Personally, I think it's trivial.

I

Meaning, I think it's done right. um There was a comment: yeah ned.

E

J

I

All right, if you look at ned's, comment scroll down a little bit scroll down more scroll down more okay and stop okay, um you can see uh how he ends his comment there. He says if b and c are not hardware only a then I agree with dave's analysis. Okay, and you can see, I respond. The text says the device illustrated includes a a biostored and read-only memory b, an operating system, kernel and c.

I

An application or workload so b and c are not hardware by definition, because b is a operating system kernel and an application. So, since his, if is true, then he agrees with my analysis.

D

Okay, all right, so what you're saying is that he didn't read it deeply enough to appreciate.

I

He read the tips he didn't read the context around the diffs, because the context around the dis answered his question. I didn't have to change that because the text already said that right all right.

D

Let's merge this then, as it is then.

I

Since he's not here, then I don't know that's why I think it's trivial, because.

I

All right, okay, he's already got two other approvals, and he said he agreed. If that was true- and I just showed it was true of the text.

D

All right, so this is uh very many um diffs um there from hank going through and just editing.

D

um I didn't have any great problems with anything here.

I

I started, but I didn't get very far. I added two uh suggestions, but then I still I've only gotten as far as my second suggestion. Okay, do you want more time to look through it then either on the call or offline I'm happy to do either one so.

B

D

There's a suggestion.

I

Or I'm happy to go away and come back in 15 minutes and see if I'm at the end. So um now look here yeah! So yes, hank isn't on right other places. He has drawn.

C

I

Some procedures.

H

I

Yeah, either leave the a or add the s. I don't care which.

H

Yeah so you're.

D

B

There could be an s here.

H

The the uh actually the plurals.

I

Okay, so you're saying do it the other way hank. You can see my suggestion.

H

I

The a back which you.

H

Want to be sorry yeah, but it's consistent. As I said I I was. I was moving a lot of stuff to plural.

I

H

I

Is now they're.

H

I

To um michael because s has to go on procedure and not attestation, yeah.

H

The next line, not in this line, michael.

H

I

Exactly yeah, I agree.

I

Okay, sorry, I interrupted you hank, but it's good so continue.

H

Because we're talking about plural here, uh policies and policy uh is still, I think, a little bit uh mixed today uh to some extent to due to my fault of not realizing that uh policy is actually okay to use. uh So I may I might have changed my mind midway on this. Thank you yeah. So, but just a set up. It's the only thing, I'm a little bit uh sorry about um that did not really benefit.

H

I

Michael add suggestion to batch, since I think we all agree on that: one: okay, uh next one, english meaning, I think, if not is better than alternatively, uh because it's talking about and then it has to decide it's not an alternative to deciding it's if it decides not to then, and so that's why I changed. Alternatively, if.

I

Not that's fine, thank you and that's as far as I've gotten so far as to that line. So.

H

Again, it's supposed to be editorial to some extent of the composite device. I changed.

C

H

But I went through a way with that, because that was his original text.

I

I did have another comment later on. I think I had a third one, so I think I must have gotten five there, because I'm not remembering a third comment um and it was, I think, in one place, you'd edit in a remote attestation, as opposed to remote attestation procedures like you're, adding the word remote, but not procedures, and I didn't know if that was intentional or not intentional.

H

That might have been intentional. That's like this! No.

I

There it is there's my question right so compare 333 and 337.

H

Yeah, that's a good point, so it could not be at the station and I was not sure because remote attestation procedures of course could be also always interpreted as rats and I'm not sure.

D

If we should have a floor.

H

Makes use of rats I'm not I'm pretty sure they don't. So I didn't want to create that association. There unintentionally.

I

My only point is, I was expecting three three three and three three seven to use the same term whatever it is, unless you can say, okay,.

J

I

So I don't care which term it is so, but it seems, like consistency, would be good.

H

Yeah consistency is good, so yeah small non-capitalized remote station can become remote station procedures. Capitalized here I agree and somehow this is not procedures.

D

I

D

Sorry, oh too many e's.

I

Yeah there you go.

D

I

Okay cool that one's as far as I've gotten, so that was my third. I.

D

Thought you were saying asking if I should capitalize the trailing s.

C

I

I wasn't gonna ask that if.

H

We could slowly run through this. I don't know if this is the best.

I

Yeah, I I said: I'm okay either way. So if you want me to review it offline, I'm guessing it'll. Take me another 15, 20 minutes or something to go through the rest. All right.

D

What I would prefer is that we spent the time for now on the issues or the other pull requests. Yeah.

C

So I'm going to briefly.

D

Go and look at lawrence's rewording.

I

uh There, I think, both hank- and I had issues with this since as being a meaning change that affects a lot of this document and multiple other documents. If we were to accept this.

D

Yeah, I couldn't really. I couldn't really understand the art, the um the right, the right word, the um motivation for this, uh except that I like blank lines, so um shorter paragraphs, I think, are easier to read. But beyond that I couldn't tell the text just seemed to move around on me now. The meeting.

I

Change is that previously your evidence was, it was generated by a tester and a tester can have multiple layers of you know: testing environments to target environments.

I

Okay, he changed it such that each layer is called a separate, a tester, and so that means you either have a whole bunch of evidences in a chain which you use that term or you have one evidence that cover comes from multiple attesters, which is also you know, problematic in other texts and things, and so and because he's talking about the root of tester and several other attesters that are not roots and said. No, that's not the term that we've defined so.

I

I think that's the same thing. You were calling out too hank, yes, yeah, that is, that is like they changing the concept yeah, because.

H

I

A lot of other places in this document it would change each, it would change the type document it would change the ccc white paper and so on. So I think we should not go here.

I

The ripple effect as hint calls. It is true.

D

Is it claims or evidence.

I

Well, I can eat, you, have multiple claim sets in the evidence right and then he would have redefined it to say each claim said and the evidence is it called. You know. Maybe.

D

Evidence right so I think that's big enough to given that he's not on the call. I think that's big enough to just um push back and say um there. um Sorry, let me actually ask uh say this change would have imp implications to many other documents, t uh what else did you say: eat eat.

I

D

I

And other sections of this document.

I

And there's probably more but at least out of the.

D

I don't know which one they're working in your.

I

Pocket, let's, let's.

D

Leave it with that that I didn't feel I had a strong reason to change it, um um so I'm just wanna uh leave it like that. Okay,.

D

That was the whole document. Okay, so what I would like to do is hank. Do you want to lead us to some of the some specific ones that you think are a bigger problem? There was also some from dave and yeah.

I

The bottom two are ones that I filed like right at the end of the session or something based on discussion.

D

uh Hank, do you want to lead us to anyone specific or just should I pick one at random.

H

I would start at the beginning, so to speak from my from the scope of my that's not the beginning. That's the end um so um yeah that one it's simple stuff, but it was beyond editorial. I guess.

D

These are comments from catherine. No, these are comments from me. Oh okay, I thought you were you.

H

D

H

For review, if you look at these two sections, you see a strong asymmetry between those two, um especially because reference values, talks about format and conveyance a lot and as a conceptual message. Okay, but appraiser policies doesn't it is not a conceptual message that we cover in this document.

H

Maybe that's the reason, but uh when you read it the first time and you don't know that, uh I guess when you read it in the same context or the same weight in this section, this in the super section uh it to me, it was surprising that uh 3.1 was so short and and somehow and suddenly I read a lot of things about how to convey reference values in 3.2.

H

I was like okay, but I didn't see that in 3.1. Why so? That is a surprising to the reader and violet's principle of this surprise.

I

um uh I'm surprised as well I'd have to look at the text to see what you're referring.

I

To meaning after reading it, I will probably agree with you, but that doesn't really falls to me.

D

D

So that's the two bits.

H

I

Is that the two bits that doesn't look like it is it is I reviewed now? I I see exactly what you're saying now, because you're referring to three bullets being in 3.2 and there's no equivalent of those three bullets in 3.1 and that's the asymmetry that you're observing.

H

Yeah probably, but it looks smaller than I remember it, but that's probably just a screen here.

I

um I agree with your.

C

Issue, I don't necessarily.

I

Right now make it better preferred way to fix it, but I agree with your issue.

D

um So so I think that this comes partly because we are. uh We believe that appraisal policies are primarily um things that you type in an operator types into the thing.

I

Don't it's tough to get conveyed across a line in the data in the diagram, whether they're conveyed across a cli or they're, conveyed across a protocol in the implementation that I'm familiar with? I'm not the implementer of it. That's just the microsoft one. It's conveyed across our rest, api.

I

D

What I was trying to say is that that reference values were, we were more anxious about their conveyance than we were, the policies which maybe are more obviously, um security, critical they're, both equivalently critical, but um the policies maybe feel more. You know it's something I do even if, if it's a cost of, is it across an administrative boundary.

I

I

So I think the points of the three look I'm not reading through this again. I think one reason for the asymmetry is about the can. The the you said, there's two things about formats and conveyance. The conveyance, I think, for the three bullets is really talking about how reference value providers.

I

It's really an observation based on the fact that reference value providers can easily be combined with other roles more easily than say. The uh verifier owner could be uh combined with some other role, uh and so that's why you know the the conveyance itself may be combined with one of their lines. I think is the point of the three bullets there, where that's probably much less true in the appraisal.

I

I

But then your second one is about data format and that's the bottom one where it talks about the actual data format. Our specific implementations doesn't define a general purpose, format for them or general means for comparison.

I

We could have a similar sentence about the actual data format and is up to implementation. The architecture does not define any general purpose format for them. We could have an equivalent statement to that in the appraisal, policies. That would be pretty easy to add. It's the bottom equivalent.

I

So I'm thinking reading through this, I understand that the the reason for the asymmetry for conveyance makes sense to me. The reason for asymmetry in the data format doesn't, and that part is easy for me to fix or for you to fix or for any way to fix.

I

Okay, what do you think hank.

H

Yeah, I'm I'm yeah. So, first of all, I I would say uh we want to want to keep people in the dark with respect to why we think it's less applicable to appraisals, so that should be content, probably in 3.1 that prepares uh 3.2, which okay and this this makes more sense here and that's because in 3.1 it didn't make so much sense, so the reader is not so.

H

This is basically understands why this document does not really talk about how these move from entity to entity how they collapsed, a movement between collapse, entities and such so I'm fine with with, if there would be something also addressing that uh with. If the format thing yeah, I would agree, is easier to fix.

I

um We could also say less in 3.2, because I mentioned those three bullets basically just say to me anyway, that your conveyance could be combined with something else, and we could collapse those three bullets into something less precise and just say they could be conveyed as part of some other message as well. But that loses a little bit of precision.

H

To this message type here, but I think down below where the bow tie diagram lives. um We uh we elaborate on on how to leverage protocols and convey things performance a lot. I don't think that has to be discussed here. To be honest,.

I

You mean the three bullets you're saying.

H

Yeah yeah in general, uh also the how to how to stack stuff into existing protocols and stuff like that yeah.

I

So I would be fine if we remove the three bullets or replace those three bullets just with a single sentence, whichever you would prefer.

J

Can I interject something here so please. um uh I think that it's kind of a missing a key point in the the difference between the two of them, regardless of where an appraisal policy comes from. It's really up to the administration of where the appraisal is doing to make a decision, whether to put that appraisal policy in place and the reference values are coming from someone else and you you may be using them and you need to have them convey to you in a certain way.

J

But ultimately the trust is coming from the provider of those values, and so the uh there's a trust decision that says, I trust the provider of the values, but I have to. um I need those values in order to do the function of appraisal right, and I don't really see that in those in that there's a there's, a different kind of trust. There right. The trust for appraisal really rests on the configuration of the appraiser of what policy they're going to use, and that comes from the administration of the appraiser, the reference values.

J

You can say, I'm going to use them, but ultimately those values are no good. Your appraisals are no good.

I

um I have a different question to ask, but it's triggered by this: it's not the same topic. So, let's see if anybody else has any comments on this before I bring up.

D

I think I agree with with peter, particularly in the business the fact that you have to you. They cross a administrative boundary and that's why I think we're more anxious about them and why we have more words about them. um I don't object to making adding words to appraisal policy, but I think that removing them from reference values is not the right approach.

I

So, who wants to take a shot at uh taking this one.

I

For generating pull request, I mean.

D

So I guess we need to decide on the approach. Is we adding words appraisal policy? um I.

I

I think one simple thing to do is to add a text into appraisal policy, that's equivalent to that bottom tiny paragraph. Although it's going to be shorter than that one, because uh you can strip out the things that don't apply.

I

um But as far as what to do the three bullets, I don't have a preference.

I

Let me get these nice yeah, so not all that will apply so things like uh the reference value. Doesn't the general means for comparison? um The appraisal policies text already talks about comparison so, like that phrase, doesn't make sense. So a subset of that text makes sense up above I'm happy to do that part. If people want it, but uh I don't know what to do. What's top part.

D

Copy link address.

D

There we go, apparently we can easily refer to the paragraph itself.

D

So yeah so you're proposing this sentence. Something like this sentence be added, but that's.

C

Just something yes, I'm saying, but.

D

You didn't know about what to say about the points. I don't think we need any points. I think that's enough to do.

I

That so, michael, are you proposing no change, meaning leave the bullets in the bottom one and do nothing on the top one. For that point, no I'm suggesting.

D

That the actual data format this kind of sentence- I think that you're saying that we're a saying we're not telling you how to what what they look like and we're not telling you and we're telling you it's out of scope. That's what I got from this last bit that I think you're saying needs to be all.

I

Right all right, remember, hank had two points. One was about, the conveyance ones was about the data format. My proposal is about the data format. The conveyance is the discussion on the three bullets. Do you have any proposal for what to do on the conveyance discussion or to leave it, as is.

D

I don't think that we can say anything intelligent about the conveyance of appraisal policies. Okay, okay, so your proposing to leave the asymmetry is what I'm hearing yeah. So I don't mind that and that's- and I said that I think the asymmetry is because peter said reference.

C

D

Are crossing a boundary and appraisal policies mostly are not if, if they do enter the verifier organization from external, they enter through a legal contract um and and and and then the verifier organization has to figure out how to implement them, whether they deploy them via rest protocol or cli is not relevant to the lawyers that wrote the contract right.

D

That's what I think right, whereas, whereas I can see you know a verifier reaching out to a manufacturer and getting the reference values or vice versa, and that they would need to agree on a conveyance and uh security and blah blah blah right.

D

I

I

Any other opinions before I bring up a topic, that's related to the.

D

Data format text: well, we don't have a volunteer for this yet so I guess I'll assign myself.

I

Okay, that's fine, uh like I said, for the bottom. One I'd be happy to do it. I just didn't know what to do about the top part, but if you want to take it, I'm happy with that too I'll, probably.

C

D

I

Yeah, okay, bring up your issue, then we'll okay by question. I was actually asked a question because I was giving um the an overview of the rats uh architecture to the confidential computing consortium tac and so thomas wassati, and I were on the call. I don't know if anybody else here was, and so um uh there was a question that came up uh and I think it was in that meeting. So thomas can tell me if he remembers it and it had to do with the the, for.

I

You can see with the phrase about the actual data format and semantics of any reference. Values are specific and the architecture does not define a general purpose format for reference values. So I'm going to paraphrase the the comment um and I'll I'll give you give the actual specific question a second here, because the reference values cross administrative domains and therefore you need some interoperability right. So if you're a verifier, you might need to get reference values uh for each uh component in that layered attestation.

I

So you got to get reference values from a bunch of different entities and now, if you think in your mind about the bow tie diagram that says about the different protocols and things so what if, if each reference values provider provides stuff in a different format, then of course your verifier is going to end up implementing a bunch of different reference value format, parsers right, and so the question that came up is um okay. Well, the eat specifies current state.

I

Okay in terms of claim, sets that have the composed current state, whether that current state is in evidence or attestation results, but its current state and reference values are basically desired. State and the reference values are specific to claims, and so for each claim you might have here's the claim. Here's the expected value and so on, and the real question was: why couldn't you use eat for desired state which would make it much easier to do comparison because you've got the claim ids in them?

I

And I because this is where I said: eats- are for attestation results or each year for evidence, and the question was: couldn't you use them for reference values? So what.

D

You're really saying is um uh the reference values. Are here's a bunch of of of claims that you need to uh um you need to put yourself in this state so that you can.

I

Your version, I your version number- is five and the appraisal policy says you know. The current version has to be greater than or equal to, the version in the reference value and.

D

And so are you conceiving that verifiers, in some cases, maybe not even know what the semantics of the of the claim is just as long as they can match them exactly yeah.

I

D

That was useful to me that.

I

It could be used here because.

D

I

Then the text that says the architecture document doesn't define any general purpose. Format and specific implications is not true, because we don't make. That would not be true. In that case, we've never taught, and I said my answer was rats- has never actually discussed that, to my knowledge that questions never come up.

D

I

Seems like that's something.

D

That belongs in the eat document. I don't think that everyone wants to have the reference values as no.

I

No, no! No! It was a good because.

D

This is saying it's outside the.

I

Scope of rats yeah because.

D

That was what I was saying before in.

I

The overview was to say, eat only covers um attestation results and evidence, potentially right. If you want to use each in in either both of those, but it does not cover reference values. The answer was well why it seems like it would make sense. What's the point so.

F

It should just say that it's going to be done by another document.

I

Yeah, that would be fine right can eat, handle, arrange nothing.

H

You cannot refer to future work. That is not a legitimate way to do this. uh Unfortunately,.

I

So eliot not at present, but it could define a minimum or a maximum. When your appraisal policy could say whether it's the minimum of the maximum, um you could construct a way to put it in an eat, but I don't think there's anything in eat right now. That covers ranges, nothing that makes it hard or sets either right jesus. You could ask the same question about set celia. You know a or.

I

Yeah, I know because the answer is no, it can just be here's the type length and value, and so the type is a particular claim and the value is you know whatever and that's what it can do right now.

A

Right code to run the question is uh what I think the overarching question is how important to how important is uh this sort of interoperability capability in terms of the the reference values in the con in terms of standardizing? Is it so important that you would want to warp the eat architecture at this point? That's the question.

H

Oh, I think that is a part of the question. The question is, do you want to impose the task of interoperability of reference values to each now? Actually.

I

I was asking a more general question: do I want to have it be in scope for rats, because rather I would.

J

Say not to do protocols.

I

But it is trying to do data formats. Okay, is it within the charter of rest to talk about data formats for reference values or not, given that you need interoperability for reference values, this is more of a rant than e, because you could choose to do it not in e. You could do it in something.

H

Of course, so so that's true so, but the the the agreement then during chartering was we will reach out how to address that. So for now the answer is obviously no. It is not in scope of the chart.

I

Yeah, so I'm just relaying, since this was a topic that the ccc asked about, they were kind of surprised that rats was not taking on and and are and yeah. I think that you think you're you're correct that the charter does not let us right now, but perhaps we don't want to paint ourselves in the corner in the architecture document and just say it's uh and not say that it is what it doesn't do. Well, yeah.

A

So I guess what I would say is also don't paint yourself into administrative quarters right if you think that this is the right thing to do in this context to reuse eats in these in this context, but to make some changes to eats to allow for this, assuming it's a reasonable thing for each to to take on you know in that document, then, then, that you know the more specificity you have. I think, in these sorts of contexts, the less work it will be later.

I

Okay, this is what.

A

I had to bring this.

G

I

I'm going to move that text up, it affects uh what that text is going to say, and I may have to edit the text in the reference value section. Two not just add it in the add equivalent text into the um appraisal policy section. So that's why you wanted.

D

To do it, I thought I think you should do it not me, then yeah, okay, um so just um so. This is what the charter says right now: uh the procedures for this activity out of scope for this working with without retartering, and so I think it's entirely reasonable that we would. We would recharter to do what to make it possible either to do it in eat or in eat, plus, plus or whatever, um because it sounds like elliot. As elliot said you we do need some extensions to eat to make it work um depending.

I

On the appraisal policy, if all your appraisal policies are either quality or less than or greater than or something like that, that only needs one value you don't, but if you need say contains in a set um intersection is non-null or uh you know that.

D

Kind of thing yeah, so there's subsets you can do immediately, but but that's not, I don't think it's I. I think we agree that we.

C

Don't go into that here.

D

In the in the charter, but I'm I I'm, um I'm not sure exactly why this matters to this text, but I'm going to let you tell tell us, in the form of a text proposal all right, because that second.

I

Sentence, I'm going to.

D

I

Remove or or union into the first one by removing stuff so about the one that says what it doesn't do.

D

Good all right, let's go on yep.

I

D

Did you want to look at brendan's or.

I

Jump to one of hank's.

D

uh uh I was gonna, go to I go, it will go to brendan's here.

I

Okay, so uh this one was in the discussion around the three methods for freshness between timestamps nonces and um that the the topic formally known as handles, um for which I think we kind of.

H

Still known uh referred to our sentence, please, we have not.

I

Since we have not talked about that issue, yet it shall be currently referred to as the topic formerly known as handles. Okay, okay,.

C

Because there's a separate.

I

Issue for that one um so brandon had asked um why local clocks are a problem right. He said even mcu's have them so uh because I would mention, like you know: nonces don't require clocks and the the the thing formally known as handles don't require clocks, except for at the distributor, and so he says well.

I

Why is that a problem, because even mcu's have a clock- and um I answered by referring to the tee use case, because that comes up in teep, which is our number one, their first in line customer for rats in the tee use case, you have to have a trusted clock and so michael correctly asked. Well. What do you mean by trusted clock? So please scroll down, because I elaborate and answer your question.

I

Michael, a trusted clock is one that the uh that any entity outside the tee cannot tamper with and mess with the time, so it doesn't have to be synchronized even as even a trusted or secure relative clock means that your rich execution system can't tamper with the clock and so example, is uh fairly common to have tes today that the only way for code and te to get any source of time, including how much time has passed since I did x, you can ask the ree and they already can lie, and so that means, if you're trying to do something like has my search expired, and you want your te to do that check.

I

You cannot do that securely because there can be a lie in has it passed, and the answer from the re is no we're still in the valid time really. hahaha So that's what you might trust. It is somebody else can't tamper with a clock.

I

And so uh right now the proposal- and this one I think, um can affect uh maybe two places in the document. I could imagine one is in the tee use case, okay, because a lot of this discussion is very specific to the t use case where it comes up um and then maybe in the discussion later on that it refers to where we said this requires a clock and it might be problematic in some use cases. We can maybe refer back to that specific use case by section number. As an example.

I

Okay as to how to address this, if people like that approach, I'm happy to do it.

H

Fine with it, I don't know.

D

So if it's the case that that tees, usually don't have clocks today, are you you basically well.

I

D

I

I'll give you the two examples that I know of between intel and arm, because I see thomas facade just dropped off.

C

I

Intel um sgx its own, it's only access to a clock is through the ree or by trusting a component called the intel management engine that um some entities choose to trust and some entities don't choose to trust on arm or rm specifies the design and chip vendors can choose what to do and the trusted clock and the arm design is optional, and so some chip vendors do it and some don't.

A

D

Just want to say that's your.

A

Design flexibility.

I

Yeah yeah, so I'm saying I can't say which one is most, whether most tes have or most tes don't, but because it's a matter of debate, but I could say a significant number: uh don't and whether it is most or not depends on whether you think the intel management engine is trusted and whether the things that you look at are things that support the arm, one which one is a higher. You know percentage which I don't know so so I want to say most.

D

Many, so I'm just thinking about about um so the ree can't tamper with, but I think the real point is that not that whether the ree can tamper with it, but rather it's consistent with the verifiers clock.

I

No, no, if you think about in the um uh uh nan's case or the epic id, and let's talk about the uh the thing formally known as handel's case right, the verifier does not need any clock. No, it does not need a clock agreed and so um the.

D

uh Sorry, that's.

I

D

The point is he's asking the reason why we would use the the okay.

J

I

I

Different examples- sorry, I came up with an example yeah think about the uh nonce example.

I

Okay, in the nonce example, the you don't need any synchronized clock, so it doesn't matter whether the tester, if the tester, has a clock, it doesn't matter whether it has any relationship whatsoever to the um to the verifiers clock right as long as it doesn't go backwards as long as it doesn't go backwards right, and so uh we have an example in the appendix as to where the verifier needs to know uh the distinction between when the evidence was signed versus when the claim value was signed, and we said an example of how you can do that.

I

Is you put the delta into the claim right, so you say here's the value I took it this many seconds ago and as long as I can trust the attester's uh clock, then of course I can trust it was that many seconds ago right, I don't know what time it was, and so uh can I tamper with that. If I can tamper with x many seconds ago, the re could change that to be zero seconds ago.

I

D

Okay, fair enough.

D

So what are we doing with this.

I

Then exactly so, I right now the approach that I would propose. That is that um I had a very tiny amount of elaboration. I'm guessing it's one sentence or two sentences max into the use case about tees and then uh make sure that the text that talks about um one of the disadvantages is uh it needs a local clock. That's in, I think, the freshness section for um uh like timestamps.

I

I can refer back to the that, for example, you see the discussion in the te use case section something that's just a backwards reference, but doesn't have the technical content in it. So that's what I'm thinking is. The technical point would be made in the t use case as an example, and then later stuff would refer back to that section. As an example.

D

There we go okay, something like that: yeah.

I

I don't know which example I'm quite right, but.

D

I mean yeah. Okay, all right. So, let's go on. Yep sounds good uh introduction in of terms in two types of environments of an attester, suddenly echoings appear.

H

J

Without extension,.

H

Yeah, I guess so, and no no, no context, yeah. I think it's true to reordering of the text. I think.

C

H

Yeah, actually, it was.

I

In three point, three, it's expanded.

H

So the references like like uh this is just thrown in here, and we don't have any references for this at this point, especially I mean he somehow gets elaborate later on.

D

You wanna have a reference here. You it's not.

H

D

The acronym's not expanded, but that you want to reference the thing yeah.

H

This is this is crucial to everything here and they just throw out a barrage of terms, and- and maybe that is not now- the user would stop. Reading this and get educates educating himself on on what it actually means, and especially the things we're not really elaborated on, like eses, which are actually not really easy to uh find a good reference for. As I let's look at that, a little bit.

I

I think that's a reasonable point because sort of thing that their sc editor would ask. Although good luck on finding a reference for bios firmware, you can probably come up with something, but I'm curious to know which one.

H

The letter- actually, I was so a catch-all thing that buyers: okay, it's in somewhere.

D

So so yeah, so in that context I would say it's enough just to tell people what it is. Okay, you know someone's been living in uh been using their commodore 64 for the last 40 years and has never heard of a bios. So.

I

D

I

The rfc editor has a list of uh acronyms that don't need expanded your reference and stuff. I'm just wanting bios is on that list. I feel doubtful.

B

Abbreviation system, I'm just going to look it up here. Well,.

A

You're looking that up, I have a question about the introduction. It's not here about the. What about the introduction, in particular, okay.

D

Of the whole document.

A

Well, in particular, section one.

D

Yeah, okay, as opposed to the introduction of that section, is what I meant.

A

So um the introduction I I might put through a pr on this, but I'm going to get it wrong if I do, but it might be a good way to start the discussion, I think there's a little bit of motivation. That's not quite present um in the introduction. It's rather abrupt, why? Why is this architecture at all important?

A

It's a question that could be answered answered first and foremost,.

D

A

D

A

You just got into.

D

Things too, too deep too quickly,.

A

Oh yeah, we're not talking the deep end of the pool, we're talking an ocean here.

D

Okay, yeah, sometimes it gets dumb to re. You know you have to explain the fact that they're computers and iot devices and stuff and you're.

A

D

I mean I don't want to go quite so dumb, but.

A

But to say you know what: why is this important? It's probably something that's worthwhile saying.

A

Mike michael, do you want to open an issue? Do you want me to, or do you want me to open it or what.

D

I'll I'll write it something here right now: okay,.

H

When you issue a pr just mention uh that is, this intends to fix or the fixes number whatever and then.

J

E

H

It's relative yeah, good.

E

I mean if people disagree with me, tell me now.

H

No, I think we uh we for the sake of readability. We reduce text here, and maybe it's not really due to another problem now. So I think balance is fine and now the time to do this high-level things yeah, okay,.

I

um All right, so I put into the chat my uh findings. um It is not currently on the rfc editor's list of acronyms, don't need expansion, but I can find uh several rfcs that don't expand it that went to the rfc editor and don't have a reference, and so I think it's an automation on their list and I think it belongs in their list. So I mean I don't think we should don't think we need to expand bios or have a reference for it. It's my conclusion, based on other rfcs that have gone through.

I

I wouldn't I mean nobody knows what bio stands for anyway, yeah. Exactly that's why I say I think that I think.

A

I

I can send an email to their fc editor to ask them to edit their abbreviations list yeah with a star.

A

Yeah, that's like.

I

It's like nobody knows what tcp stands for either, except for people in the transport area.

A

Yeah, and and syd and stuff like that.

D

And it's different than tcp ip.

H

D

Many years ago that I I needed to list both tcp ip and tcpip as skills all three of them. Those are three different skill, separate skills. That's because tcpip includes udp, where tcp does not yeah.

A

So the other thing, by the way that I'm that I'm looking at just so that you guys know is um because this is a pretty involved architecture, one of the things that I'm contemplating is and I'm reading this fresh, because I've left it alone for a while is if you have enough uh acronyms at some point, you have to have a glossary. You can't just do it in line, because people will just get lost even if it's in the, even if it's in an appendix.

D

Yeah so we did put a terminology, but we didn't put tlas in that yeah. So yeah.

A

So I don't have an.

D

Objection to putting a glossary, particularly if it's at the, if it's in an appendix just just for the tlas.

A

I mean they get a little lost sometimes, but I'm not sure I'm not quite there. What I'm saying do it? I'm saying I'm looking at that.

I

I prefer the approach where we don't use tla's ex, very much right that you might use it in a paragraph or a section where you, where you expand it the first time and then the next three times you use it in the same section you don't. But if you look in here you don't see hardly any tlas in the entire document. You can see it's michael's scrolling right now, there's none that I've that I've spotted so yeah. So I think.

D

I think that yeah, I think, yeah I wanted.

I

A good example.

D

Of one that's really hard to read because of the tlas go: look at some of the drip documents and the worst problems that that includes, you know a dozen faa or whatever tlas that uh are really poorly chosen as far as the ietf is concerned, because they overlap um but anyway, um okay, so uh we did that we added that issue for you elliot okay, I.

A

Apologize, I'm covering for my uh my daughter, needs a ride. Sorry.

D

um So that was brendan's feedback. We did that.

D

Clarifying layered attestation: I think that relates to its pr.

D

uh Well, let's open this one, I opened this pr during the meeting or something I don't know.

I

I

um Can we talk about your title and just cover the uh uh control w issue yeah? Do you want.

D

My name: do you want me to just make that a new.

I

Issue unless it's this one, but I I thought it was another one, but if it's not, then uh that would be good to see whether we can get consensus on here, okay, so uh that means at least among those who showed up to the design team meeting yeah. I think that epic identifiers is the thing that kind of stuck.

I

Epic ids or epic identifiers. That's what I heard multiple people using in different meetings as being the most memorable term,.

D

Do we have any other suggestions? I guess is the first part. I think that that there were like four suggestions, of which this one was the most succinct.

D

I think some of them had the word handle in them like epic handle.

I

No that didn't work. um I thought I entered into an issue somewhere because I remember typing in those four things, but it could have also been. I put them into my notes. Was there any other issues failed by me? No, not okay. I wonder yeah. I must have put it into my notes or something, and then I lost some of my notes because my laptop would start got crafted, but anyway um there were things like. um uh I don't know something that started with psk's and stuff than people didn't like as much later.

I

I know I know we copied them into the chat log. I think in teep, which was after rats.

I

ah So I think if we look in the chat log for the teep session, I think the list is in there.

B

I think I can find that.

I

I

Probably that second, one.

I

I guess if you search for psk, you'll, probably find it or epic.

I

um No looks like they're just using epic ids here in red, so it must have been in the in in t. It must have been the rats jabber or something.

I

All right, sorry, I don't want to take all the time just for looking up stuff in chat, but.

D

Sorry, I think it was on the the the um tuesday session yeah. I didn't make it to the moment. It was the tuesday session yeah.

G

I

Yeah psk generated by a third party. Now you can see jeremy suggested pre-shared, opaque value, uh hana's public pre-shared, secret uh smiley, meaning maybe that's a funny thing, but uh yeah.

B

I

I just can't see his.

B

I

Is about seven lines down from your highlighted yeah.

H

Opaque is something that we already actually used at some point and uh that was debated. I, like oak fake somehow in some contexts.

I

Yeah I like fewer syllables and epic id or epic identifier, epic id.

H

They're also the trump epic bar, as I'm also saying epic, not epoch, was thrown into the mix. I think by carsten, bormann and id and trigger, although trigger, is only cool. When you look at the thing itself, when you are distributing as a trigger distributor. um Well, you know so uh um yeah, but epoch id and epoch bar. I.

I

Think is there anybody that doesn't that they would object to use the term epic id or ipakaiji, depending on which way you like to pronounce it, because that seems to be the one that people were the most or the the least disenamored. With I mean the.

H

Yeah, I I don't agree with the problem at first, so so that is a problem. I think I don't think that handle is overloaded, just because something it's called something handle it's an epoch handle and I think it's qualified, and that is not a problem that.

J

H

First answer: to be honest: yeah.

J

Yeah, I think I heard that I don't have a problem with the word handle.

I

I prefer the term epic identifier over epic handle.

H

But this identifies fine id could be both identity or identifier, yeah. That's all.

I

Yeah, I agree with you there, epic identifier so far is I mean, even if I didn't have a problem with that with uh handle. I thought I still like epic identifier better. So if we were, if we were voting, which is it contracted.

F

I

F

But is it contracted.

I

F

Heck there is also um the word handles will be used by the doi system, the rs rfc. You know um I'll give you the rfc number the handles.

I

One of the points that was made that there's a bunch of other uses of handles that they didn't at least I think russ was one of the people that brought it up and other people agreed so yeah russ.

F

You'd know yeah.

H

So everything here is used already. I think identifiers also use a lot in things, so I think it's even more generic than handling, so it does not really solve a problem, but if you think that is a problem worth solving, I'm fine with that, so I'm not I'm not voting against it. I'm just highlighting we are substituting a specific problem with a more generic problem.

I

Yeah so hank I take it. Your point is you should always start with the word epoch, yeah, that's and whether the thing that comes after that is handler, identifier is not a problem. Handle itself might be a problem, but if it's always an epoch handle, then that would be your preference.

I

My preference is, I prefer identifier over handle, but completely agree with you, starting with epic in all cases, and never just using the word by itself like you would never use the word identifier by itself either. It would always be epic identifier or upper candle.

D

I guess the issue is that we have this document, which is not actually in our actually it's a informational, not a standard. um When is it cnri and the handle system with uppercase letters and.

J

No qualifiers for it yeah.

D

And it refers to uh domain name service and all these other very high level, very ubiquitous things.

D

J

I'm not very high.

H

D

This is concerned about the their word.

H

The bad word yeah, first of all, and also um there was a lot of discussion on this already and there is already a second round compromise, so we're gonna do a third round. I'm not I'm not arguing against that.

I

Okay, I'm just observing we've got about one minute left in our your scheduled time. So hank are are.

D

You gonna go with with one of these.

I

H

I am, I will just abstain on this boat. I would like to see epoch bar there because I said it and I think I have to play that for carson bowman. Yes, so um deepak barr.

D

Is that what it you said, I don't know what that means.

H

Bye, it's like when you sprint olympics and that are boris. You pass and then you'll get points and go into the new epochs, whatever it's okay. This is like something.

D

That's the thing it was just.

H

I'm just relaying it. I think it would like look weirdly at me when I don't relay it so I did, and that is it.

D

Okay, unless it involves alcohol, I don't want to talk about bars, but anyway.

D

uh So I'm assigning this to you hank what and we're out of time.

H

Thank you for the puns. All of you. Thank you. So much uh yeah, that's mine. Now, okay, I will keep it.

D

So um I'm just gonna say that that we, so we have 22 23 issues here, um maybe we'll close some without fixing them, but that implies that we have around 20 pull requests to uh generate, and I think that we really need to get this done in the next six weeks. I don't know if we need to add a second meeting during the week, but I'd rather I'd rather I'd rather have more pull requests than more uh time together.

I

I agree I I suspect it would be more efficient for us to not have two meetings each week and do the hours.

D

If you want a second hour use that.

I

Hour to generate requests yourself.

D

Yeah, I agree: yeah, okay, all right so we'll talk again next week, but let's be talking about uh specific things um and um you're gonna. I would like to do this close this one this week and actually, I think that maybe we should issue a draft with those changes just so we have a better basis to work on um so we're not tripping over um editorial changes.

D

I mean, obviously I'm going to merge into master, but I still think that it's useful that anyway, you understand what I mean. Okay, wonderful talk to you next week.

H

H

H

H