Internet Engineering Task Force Interim Meetings, 19 Feb 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SUIT WG Interim Meeting, 2020-02-19

Description

SUIT WG Interim Meeting, 2020-02-19

A

A

How do I make that go there? All right, don't need Hank on the screen.

B

No just hanging, oh.

A

A

Are they up on the shirt, oh yeah? Okay, then, let's make sure I can grab them. Otherwise, when it's time for Manuel to come up here, have them on a flash drive just in case I can't get them.

A

I already have Brennan slides, open, okay, the manual.

A

Anybody else that has slides besides Brenda and bring them up on a flash drive. According to my computer yeah, if you need a flash drive, we have flash drives all right. So let's go ahead and get started because we're already recording.

A

Alright welcome this is now an official IETF meeting. Unlike other restaurant, this is an official IETF meeting. Welcome to the suit virtual interim. This is being recorded, because this is an official ITF meeting. This part is actually covered by the note. Well, is everybody familiar with a now our new people here at the hackathon.

A

Okay, if you haven't been through that I could know well before I. Just means that, if you're in the conversation and talking about stuff- and you know that there's intellectual property associated with your our Bob better- to disclose it with in a timely fashion during the discussion or shortly afterwards for some definitional short, that's basically what this says there's the legalese on there. I am NOT a lawyer, but that's roughly what it means is. It has to do with intellectual property rights all right. We need a note-taker now, of course, the slides.

A

We have all the notes in it, and so the job of a note-taker may be really really easy. If there's additional discussion that comes up, we need someone to be able to take notes. This goes into the Proceedings of the IGF.

A

This is where we pause and we do not go on until we get a volunteer. It should be somebody who's, not gonna, be up here presenting. So it is somebody Brennan is off the hook: okay, Russ, okay, great!

A

We are not using jabber or anything because WebEx takes care of Obed. Okay. This is our published agenda. It was published out ahead of time among the hackathon, those of you who are here yesterday, if you're recording. Then yesterday we had a report out halfway through the hackathon of what each of the soot and rats and teeth groups did this part of the meeting we're going to have the soot report outs after the virtual inter meeting is done then we'll do the teeth and rats report outs.

A

We won't make the soot report outs happen twice right, but for the next hour, or so we can dive into any discussions around the soot documents and so on all right. So with that, are there any other agenda things I know the presenters, Brendon and you're gonna present, okay and for Rodney? What's your first name? Okay: okay! Is there any other presenters besides Brennan and Casper? This is Jenna fashion portion.

A

If you have report outs at the end, you can go after the slides, that's fine you'll just be on audio, and it just means the note-taker will have to do a slightly better job at taking your section. I guess, you can't just say, see: slides, yeah. Okay, then, with that I think I'm ready to come up Brendan and walk through those slides. This is any questions.

A

See yours is that one right: okay, there, you go all right. This.

C

Is the mic for the recording, so don't stay next to the mic? That's it I think I can.

C

Alright, so one of the first items that I've done with, hopefully folks here, what's necessary in team in order to sorry what's necessary in suit to support team. There is an open question about exactly how security domain construction should work if suit is being used as the transport for tea and the illusion that we've come to is essentially what I've got here on the slides, a specific ordering of fields in the component ID, it was a hierarchical view of security domains, the authorities that create them which tee easily run on and what the TAS are.

C

So we are hoping that this will be adequately compatible with most tea implementations. If you have any feedback on this, that would be very helpful on the corollary to that is that we're assuming lazy security domain instantiation, as in when you access a particular component ID, which translates to a particular security domain. If that security domain has not already been created, it will instantiate it when you attempt to bring something into it.

A

So that here in the back, you know you can't necessarily be heard so here. If you want to ask a question, this is for TVs. Did you not have a concept of an explicit security name like s GX, we should be clear when putting in the Texas to whether the security domain level in the path is actually mandatory, in which case you have to put something in there or whether you can you hide it when there's no such thing yeah, it's an excellent point. Thank you. Feedback.

C

Already already.

A

For the people in the room, yes, I'll hand, those two I might suggest keeping it in the case. If there is no such thing as yes, because it makes your message your order, yeah.

C

I would agree and I mean for I'm a super suspect, I'm disfavor area, the the suit processor shouldn't actually care too much so I'm not worried about that, and so there's still a missing slot for Tammy rice in the suit draft.

C

So I think that that's an extension point that needs to be added into the suit draft for publication, since I believe that is now the only remaining element that needs to be covered to support teep directly in suit and I think there should be a TA example in the suit draft, so that some of this is clearer.

C

We also did a bit of work on integrating suit and rats, and what that's going to look like is effectively defining some a place for suit specifically now. The way that this should work in suit is that the the manifest should be able to define which elements are going to be added to the attestation report, and so essentially, eight claims will be appended to the attestation report on the as the manifest is processed.

C

So there's a few things in there that we'll need a little vacation to the manifest, but that will essentially be replacing some currently nil arguments with a field that says yes, you should have test this.

D

C

D

Is Dave welcome, IRAs chair, so did you guys talk about where this work would actually happen?.

C

Sorry seconds and we were worried so I'm I'm, open to suggestions on that. If it's better to be done in suit, that's fine! If it's better to be done in eat in rats. That's fine I think that it probably should be a separate draft from the suit draft, so I'll put the plumbing in or a suit to recognize what should be attested but I'll.

C

Do it in a generic way so that it's not specifically to any other to any other draft, and then that information can also be used in rats or possibly in a secondary suit draft to define those eat. Claims.

D

Participant I, don't have any personal. You know views on where this work should be done. I just wanted to point out that we would have to update our Charter if we were to take this work on inside.

E

And it should not be done in the document yeah, okay, because people may be stopping the progression of the manifest graph due to the dependency to the each raft. So we will do the rest, but in a separate plane set. Yes.

C

Okay on to the next one, there is an open question about the, including of the authentication wrapper is currently the authentication wrapper contains Cosi objects, the authentication wrapper itself is b string wrapped, but the cose objects inside the authentication wrapper or not, and there's a question as to whether B string wrapping them would make it easier to integrate with existing Cosi libraries. So I would appreciate feedback I've gotten some feedback already. That's suggests. This is a good plan, but I'd appreciate any additional feedback that on adding that to bite.

F

You have to hash it right, that's what Lauren's going blade! You have to hash the objects that you're good.

C

So now this this is rapping the signing objects themselves, I'm gonna be string, and the question is whether that would be helpful to exist in cozy libraries, no.

F

Opinion on that.

C

Okay, so I'll leave that as an open question. Next, you start a thread on the list. Yes and anything, that's in these slides will presumably I will. If there's not a clear resolution, I'll be opening a thread.

C

A

Have time between now and ihe after pokes me works on the draft and so getting planning to be yes, so the point is for any of these things, even if there's no discussion on the list after posting the thread right does, maybe nobody really cares either way and either answer is fine, pick an answer put in the draft yep and then we can report it out and verify that nobody objects, then excellent sounds would be great.

C

So there's a question about putting the question: I have taken the decision to put a digest of the manifest in the cozy payload section, and the reason for this is to enable modular processing of large post quantum crypto signatures. If this is a bad assumption, I'd like to hear about it, but for now this is how it stands. There's also a question as to whether it matters if you're using EDD si, which in the cozy draft says for very large payloads.

C

You may want to use a hash and transported separately, which is what I'm doing so. This might be helpful, I'm, not sure what extremely large means in this situation.

E

Hank at the mic, so I think that this is in general, a good idea, because we are thinking about having this just momentarily called the out of wrapper and could maybe the sweet envelope and have multiple things in there. And sometimes the same should be discarded by the constraint device, and this is very easy to discard. If you just have a digest, referencing it from the sign manifest and not the whole thing in the manifest that is very hard to cut out and still serve alidade at them.

E

So I think that having this digest through a payload feature in the envelope in general seems to be a good idea, independent of post partum and en ESA incursion.

F

Isn't there an extruder there.

C

Absolutely is, and that was the original use in suit. It was just a question of whether external data was necessarily compatible with a post quantum signatures. So the issue here just bear with me. The interest at the the point here is that you would have to first load the the manifest portion so that you could digest it so that you could do so that you could verify that digest with the signature which is behind, which is which you've now skipped over. So you have to seek backwards, and then you have to do that.

C

Verification and you need to use it. So you have to seek forwards again. So this was what I was trying to eliminate. That's where I come into the modular processing question.

C

So next it reference URI talking about using claims. It makes a certain amount of sense to be able to say this is where the canonical version of this manifest is stored. If you discover that you don't have the whole copy that some field has been discarded, this is where you should go to get a copy of it, maybe a template so that you attach a hex encoded version of the digest to the end of it or something similar to that to reduce the transmission size.

C

If you're going to attest the URI and the digest, it doesn't make sense to do that. Twice. That's open for discussion, but I think that adding a reference URI in general is probably the right thing to do as an option. No I think there might be some some use cases where it's mandatory, but I think that as an overall concept, it's optional the examples there are some errors: the vendor, ID and class ID per our parameters in the CD DL, that's published in the draft. They are parameters in the description of the draft.

C

However, in the examples they are not, and so that that is an open error which I'm going to fix.

C

Likewise, the try each examples do not have a string wrappers around them they're supposed to, but they don't so I need to fix that thanks guys, it looks like minimal loops are going to be a benefit. That's something that we need to look into. Specifically the loop allowing you to do the same operation for a list of components appears to be a very useful optimization.

C

Now the the driving use case for this that made me believe that I actually really need it is that a digest would have to be repeated multiple times if it were clocking between components and needed to be verified after each copy. That seems like a problem so, rather than having it duplicated or having some mechanism to assign from one parameter to another, which is currently not an option, this would allow you to do the same operation to a list of components so assigning the same thing to multiple places works in a fairly straightforward way.

C

So far, I haven't found a compelling use case for map test. Execute, which is something we've discussed in the past, except for the prioritized parameter lists that I have been asked for it, but that seems like a pretty nice use case so far, so I'm not exploring it at the moment.

C

It's another minimal loop, which is that was my last slide, but I have one more thing to add. It appears that the Seaboard, encoding or decoding that I've been doing is going to be pulled out of my library, maybe available more widely. It's a don't know if that's useful to anyone. So just not mention it. Wait.

A

Like I wanted to record it here, the discussion we had at the end of day yesterday, which stemmed from people back your slide about the minimalists loops yeah. Yes, so the question had to do with even have this line about the suit aired and recorded in the president, I.

C

A

If you can talk about that a bit- and there was the question about whether.

C

We would go past the first errand, okay, okay, so so the idea that was discussed was the into the manifest is almost or is necessary and almost sufficient to indicate where an update has failed and so adding that as an additional attestation claim might be interesting. It's not necessarily a measurement yeah.

E

C

It is effectively a measurement, so the idea here is that a device would report probably two pointers because of this structure, so one pointer would be which component ID it was on and the other pointer would be where it was in the sequence and.

A

That would be sufficient as long as you bail out of the loop as soon as you hit the first error. That is.

C

There's only one.

A

Error possible at a time.

C

A

That is sufficient information. There.

C

Was another question as to whether conditions that have failed should also be included in the attestation list? Now? My concern with that approach is that it could get quite large in some cases and I'm not sure how to handle that one yeah I haven't thought about urban I should put welcome. Is it for me.

A

Why so are these slides been uploaded yet by somebody else? Okay, great so a day for anybody remotely, they should be up shortly. Oh interesting I have to unplug this so that I have space to fit in the USB Drive. Those are the ones are too close together. Okay,.

A

Can you just use that slider this fit I, don't know how to get to full screen in this? Is that sufficient for you? You don't know if this will go up and down now you'll have to you have to use page up page down on the keys here. Oh I know I could say this isn't doing because I unplugged it yeah.

G

A

G

It's okay, great she'll, be put on what we did with implementation. Suitable addition in Wyatt. Where did we start? There was already ignition that, based on from the last year's first.

G

So you can use it, but it's the old version of the draft, so we tried obviously to update it to the new strap. Succeeds. We nicely suit into a good system everything we refactor our code a lot at a unit test, so we have that and then we made like having of sale on the new draft structure.

G

We were told like there were not so many changes. We got to know that. Well that so many check is PC, so yeah.

G

So what do we learn? We found has been said. Some glitches meeting ran was very fast to fix, most of them already I. Think all the other pics ended the cute yeah and realized that the structure of the manifest is quite difficult. We asked for some graphical representation random. We discussed tools so yeah.

A

The next presentation doesn't have any slides, so we can't talk I.

H

Do have a flaming I did write this up on a wiki page for the hackathon link to though I.

A

Don't remember what they would be pages is linked from the it's linked from, though.

I

H

Left takes you one down, but on the right, the webpages it's hurt by that, but the home which we needed. The third.

A

But it wasn't choice: MC butea,.

H

Alright, so I'm David, Brown I, think everybody here knows that so our goal for the the suit hackathon started out as incorporate Brendan's food parser into MCU boot turns out. That was a bit lofty of the goal for the hackathon we produced this go a bit and we're almost successful with that.

H

So what we basically ended up doing is we've modified the tool in MZ boot called an image tool to accept a suit argument on its sign command, and this invokes the Python manifest generator currently with a hard-coded manifest that has nothing to do with the image or signing generates this hard-coded manifest. It then signs it with the Keen. That's in the MCU moved and then embed that in the image in place of the manifest in MCU boots bespoke format.

H

This then within MCU boot itself, I, basically stubbed out the row that does all of the existing verification. I find this manifest copy it to ram, and then I jumped into the suit parser the initial function in the suit Parcher, which basically just walked through it and pulls a sequence number out of it and when we're wrapping up, we just got to the point where that worked. So that's our status, we can. We can print them.

H

Yeah so that I mean the complexity here as far as what comes next, the MCU moves.

H

Logic is a bunch of C code that does a bunch of steps to find the images and then decide what to do would then may do a swap it maybe will run with the suit manifest some of those things or instructions in the manifest itself. So there's a bit of unweaving of the MC boot code so that these things are morning functions that would get called some code.

H

That would call those based on what the manifest says, but I think this is I, sent a little status, update to the other maintainer of MC boot and he's like. Are there four requests ready, and so there is definitely some interest in this. So we'll move forward with this any questions.

C

Not the mic is there. Anything I can do to make sure that updates that are done to the suit loader code, especially in regards to the examples that don't match the CD DL makes it into the work you're doing.

H

Yes, as long as I know that they happen, we are a sub-module but hit sub-module. The code is being it's probably gonna end up having to be a fork if I have to modify it first to make it work, but currently the at least the parser just compiled without changes so but I will be in communication with you.

D

All right thanks.

A

Anybody else from the suit group that wants to report out on what they accomplished. Besides the three that already in there is everything kind of summarized in those three and you know Brendan, you said the C forum was going to be separated from your implementation into a separate piece so that people could decide whether they're like if the discussion was, if you're only use of C bar is for the suit parser, then the 250 bikers might be sufficient.

A

You also are doing up he's with Seaborn. You probably have a full C where a parser for your other purpose and you can use that with a suit bed rest. So that was the discussion we had all right. Is there any 278? Yet? Okay, well extra 28, bytes, sorry, Brendan you're out there other things that people wanted to bring up, discuss.

A

This one use the super, sir and yeah. If you want to come up and talk about this one, okay,.

J

Your name yes I'm huge sector in this hug song I tried to implement my suit period. A show current implementation, Lonnie on bhakti environment first I, checked. Interoperability. Is that my suit us latest my question right. That's that is okay. Second I built.

J

J

Up I'll go with man, crest.

J

J

Banking up past and that my fist, but what you have to clarify my face.

J

Sui-Lan in this hacksaw.

J

My first time I touched development.

J

Specifically complex big files so.

J

Implementation leaves a few external rather so.

J

Maybe number Webster library, Greece's I, think T a beaut.

J

J

K

On te site in opti, did you does that run? Is a trusted application or inside walking? Yes, just application inside.

K

Regular pH or acidity.

B

A

You anybody else, so I can go back to the wiki here.

G

Oh question: not just what I mention it but I know that say.

A

Mrs. Araz this one please come up and say something yeah you left already, but no that's not worth it. Okay, gosh yeah! You.

I

Can talk about.

I

A

Is it actually they're.

B

Actually things no I.

A

Wonder if the link change, let's go back and just nobody Frank.

G

Before we just refresh.

G

Anyway, nested does a nice message basically managed to package provide for both tea, cozy and cute super juicy person fancy, and we certainly welcome a purpose from I watch it on compiling dings, fine.

A

A

A

And this was from I, don't know how to say his name, but he left right thanks for reporting up for all right. Are there any other report outs from the hackathon learnings, as I mean from the suit I got that nice? This is a question to see everything covered great seems like we've made a bunch of good progress, so I guess next steps and the action items there's two things there's what's on the drafts, I think within the suit working group.

A

The only thing that I've heard that needs a draft update is the suit manifest document. That's what Brendan was talking about. Is that should be done before the internet draft deadline, which is like second week of March second week of March a so, but not that one and I didn't hear anything that wouldn't necessarily need changes to architecture information model separately. There would be a suit related document in the rats working through that might be a new, a new document.

A

That's what I heard and we should verify with the chairs and stuff that we can add a new document to there as opposed to adding it to the say that the suit working group charter, and so here at the inclination of the people in the room here. What I've heard is that it's better to do that in rats right, but we need to run that by our ATS I expect. That would be fine.

A

So then separately we have a hackathon coming up at ATF, 107 and I hope that a bunch of this work can resume there at a suit table at IETF. One offset yes, how many people here gonna be at the IETF 107 hackathon raise your hand, Vancouver Vancouver, and it's the what's: the March 21st through 26th summer and at 7 that every so I see and so I count one two three four five actually I'm gonna count twice raise your head.

A

If you're gonna beat the hackathon and I'm gonna, have you put down that I'm gonna say who's gonna get the suit table right. Okay, so right now, if you're just gonna be at the hackathon, raise your hand for one two, three, four: five: six, seven, eight nine and eleven people in the room. I, don't know if there's anybody remote that wants to raise our hand. I kind of eleven in the room gave you so they're welcome here, I.

D

A

Are you gonna meet thanketh on yeah, okay, twelve great, and let me know if there's anybody else, I didn't see anybody else on WebEx, yet so all right, how many people would be sitting at the suit table at hackathon, as opposed to just present in the room?

A

Okay, some people move back and forth between tables yeah, okay, I see at least three people in the room that would be at the suit table. Some of us would be in an adjacent team when on call back and several people in that all right. Thank you.

A

David Wolter, ever since I've been stable thanks. Okay, is there anything else that we have before we close the intermediate.

A

We're recording it as part of the proceedings of the discover by node well right, and so we don't have to be. We will start. We will stop recording okay, whether the web that stays open after that I don't know.

A

Yes, they are publicly accessible. So if you go to the IETF site, you go under meetings, you look at proceedings and then all the proceedings that past virtual meetings are there and they're all linked to the actual recordings.

A

They won't be there next week, but yeah I'll be there by the time we get to IETF 107. Okay, then, if there's nothing else, you have anything else: okay, good! Let's go ahead and close the meeting and hey Walter any last words before we close.

D

Thank you. Everyone.

D

A

Just curious, how was the audio for your remotely be able to hear it.

D

Wasn't bad I mean there were a few like muted conversations in the background, but it was mostly good. We've.

A

Been having internet connectivity issues all day, so we're happy to hear that the audio was reliable when we're scared that it was gonna be flaky, so we surprised all right. Thank you. We will go ahead. Thank you. Bye, bye,.

A

I recorder or save recorded meeting do I have to save the recorded meeting. I'm gonna say it anyway: okay,.

A

A

That's what I think it's doing I've never done that before, but apparently I can't stop. The recording I can only save it recorder panel. There we go stop okay,.

B

A

Think I've stopped sharing their stuff recording.

A

A

Recording is very short of just me: yeah I think yeah, and there was somebody else. I think said: okay welcome, Juan Carlos lets me to end my piece. Thank you, physical boot, the winner on the room, and then we add other people's names to list so we're just gonna head you into the blue sheet. That happens to be white thanks for done great we're, gonna sign off now. Thank you. Thanks, bye, bye,.