Internet Engineering Task Force 104, 26 Mar 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF104-ALTO-20190326-1610

Description

ALTO meeting session at IETF104
2019/03/26 1610

https://datatracker.ietf.org/meeting/104/proceedings/

A

A

B

A

Always been an excellent, no tech sure how about it.

C

A

Can take one before Sabine, while she's presenting you in, do it Richard.

A

So you're gonna take notes while she presents and then just gonna take notes for the rest greater and both of you just send me your notes. After after the session, the last times was Vijay and I just really quickly looked over them and we just uploaded them I.

A

Think personally, I honestly think the notes are not as important anymore as it used to be, because we have to meet echo recording of the recession anyway, so I mean they used to be 90 halftime, where the minutes were what people were not there, but now that you know this. Thank you zabi, and is anybody in jabber? Do we use jabber anymore, or is everything going to meet echo a million doses things there and to be standing a jabber spot because we have meet equal people can comment, but.

D

It's good to have somebody say if somebody has found video audio whatever you can still relate, not.

E

In this window, we need somebody else to watch this chatter. Okay,.

A

It can somebody watch. The chat is that the meeting will check your Java okay. Can somebody do that, so the issue is we need somebody to work. You can do it great thanks. We have no takers already, so I think we're good, and if somebody is there just yeah, so we should pass resolutions. Yes,.

A

So I think we can start I. Think we it's about time. So can you hear me as this mic on alright, so welcome everybody, and my name is sends it off I'm chairing this session here. This is the ultra working group, application-layer traffic, optimization, my co-chair BJ and unfortunately could not make it to this meeting um and I'll talk about this later. I guess, but we have high hopes that he can share a session in Montreal at the next ITF, so we will also have a physical session there um yeah.

A

So welcome and I guess everybody knows to note where basically this means there are certain routes in the ITF and you have to follow them. If you don't know that I'm not going to read all of them right here, please check them out. There are certain IPR routes. If you are aware of any IPR, you need to say so and so forth and so I'm just want to make everybody aware that we are under certain rules here, that's basically the roots of full transparency and yeah. This is actually I've review. Sucide from the last time.

A

I shared this session here, and this has not changed, I think we're making good progress, but the main focus of our working group is finishing. The Charter items and Vijay and I have been very strict about this. We are not opening any new items until we have finished the working group items and I think the working group agrees and we have been very strict and the agenda which I'm going to present in a minute has been set up accordingly. So we are really going to focus on the working group items and and yeah.

A

We want to finish our milestones as soon as possible and I do it. It has taken a while, but I do think we're making good progress. It partly is also because we found some dependencies. Richards is going to talk about that, and that was you know. What can you do if you find it find it dependency? You need to think it. You need to think it through. So I do think that people really are working hard on their working, vitals and finishing, and so yes, let's look at the agenda so basically yeah.

A

We have on the main working group items that we still have open. We each have a presentation on them. I think the only one where we don't have a presentation is the performance, metrics and I'm going to have a comment about that and Sabine. Maybe you can just briefly update us, but on all the other working items we have a presentation and I think we're gonna see then how the progress is and but we still allowed individual presentation.

A

So we have at the end of the agenda and three individual presentations that we are certainly happy to discuss. But, as I said before, EJ and I will not accept any new working group items until we have finished our milestones and then we really need to think about reach our during and probably other people need to chair the working with we do so, and this is also what Millia, what we've agreed to do right.

A

So basically, these are certainly interesting and we're happy to discuss them, but we will not discuss adopting any of these, and so is there any agenda bashing. Anybody wants comment on the agenda, anything that we forgot. Everybody's happy, okay and so progress. Yeah we had the last meeting in Montreal. I was not there, BJ was there, she chaired it. We had a virtual interim in December, where we had basically discussions on a lot of dependencies and progress on the individual documents. So we are meeting quite frequently.

A

The interim was basically instead of meeting physically at the bank of ITF and so general status. We have one more out of the working group. One more document. The cross domain discovery is nice, G pro processing and Sebastian is not here. He could tell us there are certain discussions they're having, but we're not going to discuss this today. It's in the is G and they have raised some questions again, I think also regarding the security section, and but we are.

A

We hope that this will move forward and we have then, as I said, multiple working with items which we as chairs believe are all close to being finalized. And let's see today, if we can send some more out, some of them have actually been through working class, call twice already and I'm thinking about cost calendar, and so so we can is G, and so just I mean we have presentations so at just a brief summary of our working items.

A

In my stones, cross domain server discovery as I said, is under eyes: G evaluation, we're not gonna, discuss it cost calendar and we just had an email from BJ of my co-chair today we have finished our second working plus call, and we also have addressed the issues which is checked it today and he sent an email out on the today. If you saw that he thinks it's ready.

A

Just gave some minor pointers on mostly text as some typos and stuff and Sabine, who is in the room and basically confirmed earlier in the email she's going to address these issues and then we're gonna have a new version after the ITF. And hopefully we can send that one.

B

A

Out and so we're not gonna em, so we're pretty much done with that one and see the NFC I am we have a new person published and it's moving forward and we hope it's gonna be ready soon. Jenna's got a present remotely, so I'm, not gonna or be a Richard, is gonna present. Okay, and so we have a presentation so I'm not going to say much further. The same goes for SSC. For that one we actually had a working last call and the authors have addressed this document.

A

We just cherish think it's we're polished and the comments have been addressed. Richard's gonna present today, and so that one is another one. We actually had already working bless, Kosovo close to being done now, coming to a little bit more complicated things, unified properties and path, vector, they're, making good progress. We have new versions out, but they have been dependencies and and Richard you're gonna update us on that, but we hope that we can also then I think they are resolved.

A

You will talk about it, but of course we want to move these to working class call soon. So we're gonna have asked for detailed reviews, then, and that's a quick summary and from the share perspective, is there any other business forget anything? Oh yeah, so, okay, all.

F

Right it's in China, so any sort of very loosely follow this group. Is there any quick comment you can make about implementations of these yeah.

A

So as far as I know, there are multiple implementations, so there have been presentations here from people who do alpha server for telecom. We have presentations from people that do the other server for Sprint Richard. Do you have multiple? You have an implementation going on? Maybe you can comment, you can call.

F

This is that is.

A

There, a sort like a page with a list of implementations.

F

Is there something.

A

So I would need to take through over the years. We have multiple presentations here from actual implementers, but we don't have today and I. Don't think we have an overview page for that maybe thing yeah yeah.

A

One thing you can certainly do is you can raise that question on the mailing list and I'm sure people will reply immediately and send reporters so that that I'm very positive that people who reply soon so just raise that issue in the mailing list, but Richard so yeah I.

G

Think that's a wonderful question and we had an auto survey document from last last IETF and but the workers were well now pursuing for that small document and we do have a long list of implementations and deployments and all these citations. But they were not for surely the moment and that's wonderful. So will that I think that's suggestion which we probably should polish it which should really update it will do that. That's wonderful suggestion right and.

A

It's a good point: I mean in the ITF. We also we always very keen on implementations. It's important, so yeah just raise it on the mailing list and think people will reply. One more comments. I mean go to the mic. If you have a one more comment.

C

So Sabine from Nokia as a note-taker may I ask people to take the time to articulate their name and affiliation. Thank you.

G

Can I use this chance to really because related and multiple Peter people us and, of course, I'm from yellow Sabine from Nokia and Boreas from Ericsson, and we have people from Louise from californica where people from AT&T are white people from tents, and we have quite a few people. We just of course we have Dan you from so well. We have a regular, a weekly meeting every Wednesday.

H

Actually in Italy, which is coming it that we have frequently weekly meeting all Wednesday hanging us us time, vs, Costa and please feel free to to Ewing us. Yes,.

A

As that meeting announce on the mailing list, no.

G

We don't somehow brief your that, given on your meter, we realize were mostly running is more like a maybe 10 15 H people every week. We should do you recommend we send this one to learn, meaning lists so.

A

People have been asking, and otherwise to arrange it visually. Why don't you just announce it on the mailing list that, if somebody is interested, you should contact whoever.

G

Okay, okay, great we'll do that I think both power and we saw other people. Yes, we want to have. You are not trying to be close, which is mostly just core people on computing. Maybe if you can kind of pH, okay,.

A

All right, it seems there is no other business I'm looking at meet echo, and no so that basically means we can start with our first presentation.

A

Yeah so sabine europe on cost calendar, and I can give you.

A

Click thing: let me just find your slides.

A

Cost camera so right.

A

Yeah so works right: okay, okay,.

C

So hello, this is about the progress on the auto cost calendar draft. So there was a first iesg review last december, so the review had to discuss points and but had also enough positions to pass. Once the discuss points are addressed, there was one discuss which was related to the design to question one on the format of a given field that was could cause passing arrows and the other one was like a question about some contradiction between calendars of long duration and what happens if the actual values are changing? How can the client know?

C

The other discuss? Point was more about the IPR declaration, so that has been solved immediately and so upon. This decision area director recommended that we because we had to submit a new version following the second discussion point, so we just decided to address all the review comments.

C

So this presentation therefore focuses on a discussed some discussion number one okay, so the current version of the draft is active, so it went back to the working group, and so there was a new version 10 that has been submitted in February and addresses all the first iesg review comments, and then the second working group last call ended in February 25th and there is now a new version. So this is the current version and basically it addresses like errors and JSON examples and also check the correctness of the provided example.

C

So these presentation is focusing therefore on v10. So basically the highlights on the updates. So you have updates on the design to address the discussion points. So we revised the format of the field, time interval size and also address the issue of calendars with long duration and changing values. We also referred references to RFC, and that was not like a minor edition because it was also taking into account the error of seeds that were followed by the base protocol and that are now obsoleted.

C

So there was many many clarification text that was added and section 4, 1 2, which is actually the core section of the document, because it describes also server responses with calendars. So that has been reorganized, and so we added the editorial Corrections on several formats and unit sends out time zones.

C

Ok, so as for the first design change, so the the in the attributes that div that describe a calendar, you have a member that is called time interval size that describes how long each time interval last. So we it used to be encoded in a string composed as follows. So one number and one string that was describing the time units and the first drawback, was that it could cause passing arrows. Actually, this design was inspired from how constraints, on course expressed in the base protocol.

C

They are also expressed as a string where you have like GT, 0 or LT 0, and so we took inspiration from that one. But actually we agree and we found a easy way to solve it. So.

C

So the solution is as follows: so the new format is as here as described here, so the value of the time unit is implicitly a second, so it's defined it's specified in so just.

A

To be clear, and so basically you're you're not directly following the suggestion: you're not entering and introducing two separate fields, you're saying we keep one, but it's in seconds.

A

C

The document specifies that it has to be a second, and so you just enter here adjacent number that gives the number of seconds and provided that it is required that the service same as in the base protocol. They support a triple e same five, four double precision floating point, and so, for example, instead of one hour, you just have three thousand and six hundred so and that time a duration ranges. So this is a short example again in the IRD of how the attributes that describe a calendar are encoded.

C

So the yes and the other discussion item was about the attribute repeated, so the attribute in a calendar response is a number that tells you. Okay, I, give you a calendar, and this value are good for the next and duration. If I have attributes repeated is equal to 10 for the next ten periods hits the same values, and the question was well that implicitly prohibits that a server modifies the value until the the number of repetition has elapsed. So what happens if there is a an actual change in the values?

C

So these all the questions here are listed here, so I will not read it all. I just leave it here for those who are willing to read the slides after the presentation so that there is documentation so the solution, the solution has been to add and specify things in several sections of the document.

C

Okay, in section 2, in the overview that section now explains that a calendar can be used at a timetable but like any timetable or schedule, you cannot predict unexpected events by definition. So it's just like a train table. If something happens, you cannot predict it, but at least you can figure out when you want to reserve your seats, so this section recommends that any server supporting calendars also support the incremental update service and likewise every client that uses Canada as well supports and uses that SSE that incremental update service.

C

This way, you can just get your calendars and when something happens you know it's as a client, so this has been. This has been written in section two. This text has been, of course, repeated in Section six on security, because also the security addresses the issue of what happens if the unexpected values are changing and in section seven operational. Why? Because in that section seven the document tells how how convenient it can be to use that repeated, because it saves lots, it can save transactions.

C

So we repeated this again in Section 4.1.2, we add a we added, a paragraph that explains that a server can update a repeated value, repeated Canada, once the number of repeat repetition has elapsed or in case of unexpected value chains. So this change can be retrieved drift by via incremental updates and the text. Also addressed actually another comment that was raised and with a question. So how do you deal with like in finite cannon deterioration?

C

So the response is well, you give a high number of repetition and once that time is elapsed, you just update your calendar, the other server, so they discuss to our I won't be long about it because that has been solved so additionally to these design changes. We also updated references of the era of seeds that are used. The other RF seeds that are used in this document. So the JSON format now follows RFC 8259, so the JSON format used in the base protocol used to be 7159, but it is obsolete it by this version.

C

Sorry that mandates actually utf-8 to encode text and this for a very good reason of guaranteeing a better interoperability. So this has been considered reasonable reason to stick to the new version, and so there has been a discussion among the area, directors and also in the working group. So the draft now uses the newer version of edit, so JSON encoding and the RTG has been required required to identify the extension that use UT's.

C

Utf-16 32 of any format that is not supported by this new JSON RFC. So up to now, we did not identify such extensions, and so a second update, I with respect to FC, is about the time zone that is used in the document. It's now UTC according to RFC 7231, given that it still displays GMT the time stamps still display anything. Okay, so now other hours for other update. So there has been a substantial change in section 4.1.2, but this is more if this has nothing to do with the design.

C

This is just to clarify it, because it seems that it was carrying lots of it was kind of hard to follow, and so we I we hopefully reorganized this section to distinguish in response is what members are specific to multi cast and not multi cast, whether they support calendar or not.

C

What has to be in the what has to be in response so- and this is also in this section- we also better stress the fact that the attributes that are conveyed in the response is mostly already appeared in the IRT so and there has been other updates on examples. Formats errors, hopefully everyone every error, catched and also many many valuable guidance from the reviewers.

B

C

The clear awarding Assessor! Lastly, no, this is a this- is a recap of the IES G feedback. So each time we are addressing the a SG review comment, feedback I was submitting them to the person who was review of was issuing that comment and they were agreed mostly, and we are waiting for the feedback of the application, art area and that's our other conclusion. So again, great thanks.

C

All these received feedback and guidance and the next steps we will post as soon as possible a revision upon bj's, feedbacks and qvj, and then we will be waiting for the ad feedback and is Jason for you.

D

Makin, even thank you for all these edits, so on the processing pad. We will also have to run another IETF last call just like, because that's a process right and then I will put it on a returning item on the chat. I chat and I wouldn't expect to see any further comments. So hopefully everything should be cleared now and I.

D

Have a quick question remind me: there was a discussion about if you use a new JSON format that you might also update the base ultra protocol so like that this document would update our C's 70 to 85 right. What was the resolution of that question.

C

Well, the resolution was well actually addressed here. Maybe.

A

As far as I understood it, but Sabine correct me if I'm wrong, this new document will require the new JSON format where every regarding rc8 259 and it's up to the working group to update the old yto to the the core of the protocol, because well that's an RFC and it doesn't require the new JSON format because it was not out there right. So I guess it's tricky, because we now have an extension that requires yeah. So.

D

What could what you could do also tulip bit strange? You could also just use this document to update this one specific point in the auto base. Spec then.

A

Again, I don't understand why it's tricky, because everybody who wants to implement this extension has to follow the new JSON format and, if they're smart, then they also use it for where they have it already implemented. If the core of the protocol gets updated right, it's not.

D

So it did I mean I.

A

Mean maybe you can comment at the iris juice anything regarding this, or did you just introduce this? Was this introduced after the iOS? Do that.

C

That was a I use, G feedback yeah, because somebody has to reference the new version of Jason, a new version, then I realized.

B

C

The older one was used in that base protocol and that issue discussed in the working group was. Do we have any documents that really kind of where the specification and the implementation is tied to like UTF, 7, 16 or 32 or any other formats, and actually we did not yet identify, and, yes, we'll be happy to have your guidance as well. For the moment we said. Okay, no document is tied to a format that is not anymore valid, so we just do nothing right.

A

That's what I was trying to say if somebody has a core ultra implementation, you can use utf-8 regarding to the quad instead of adheres to the coraggio specification right. So if somebody implements this extension and switches, everything to utf-8 there's no problem, there is no violation of RFC seven. Two, eight five, the Quadra protocol.

D

You're saying like if I have an implementation, it's based on a new Jason specification and I talked to a service that has the old Jason specification implemented. There shouldn't be any problem right well,.

E

A

Not that I didn't say that I said as far as I understand it, the new RFC eight, two five nine says we have to use utf-8 and the core outer protocol says you can use utf-8 or 16 or 32, and if now this extension says we also only allow utf-8. It's not necessarily a problem, because everybody who implements this extension will have to use utf-8, and if that implementer is clever, he will also use it for his core Alto implementation.

A

And then there is no issue but I think we can stop this discussion because the is G is common and it will go back to the is G and there are the Jason experts who should check we think now we are in line. You still have questions, Mia I, wasn't here: okay, yeah.

D

G

Believe this issue wouldn't apply calcio only I presume this become an issue, for example, to City I. A lot document really using generic UPF I'd be the time of most of base protocol. We, like a PID identifier. We really were explicit. You can only use a very small subset because we're trying to control the, but we do have a get on like a comments. Descriptions we just whatever Jason string, total legitimate to us, I presume those will be potential, can cause issues you're.

A

Saying Richard is we don't have as I'll to workers? Don't have any strong feelings, we're.

B

A

Advises as a general mechanism to handle this problem, which many working groups will have, we will do it and- and we fear that this is what they wanted us to do. So, let's kick it back to them and see if they agree. Yes,.

D

That's all correct I'm, just wondering what you noted insect exactly in detail and if that's the only thing you did basically is now referencing the new JSON format in this RFC right, so just to be in the same safe site. Maybe you want to add some discussion to this document and pointing it the problem that there are different Jason formats that might be used by different.

A

C

A

Can take notes a beam that also. That is something you should add to the document when you address VJs comments and not just require RFC a to five, but just explain also the implications of that mm-hm.

C

A

Then we kick it back to the is G and see what's what's happening.

D

Assistant I mean, if you have another question about this topic: I can I can feed it back to the ISD and they will have happy to help, but IRA put it as a returning item on their on the teletype. So probably nobody will look at me and kind of so it don't expect any further comments.

A

Yeah, so that basically means we should maybe take it offline to the Jason experts just to be sure.

D

I mean like there was a discussion with the ad that made this comment right so in it went for us in pack, and my question was mainly like what was the end of the discussion and I guess. He also replied and it was a discuss point. No, that wasn't discussed.

A

D

A

D

A

Sure man's opinion now this is a new version meeting with everything. Maybe she you can just send it to that ad I. Don't know this yeah again and say: look is this now how you want it: how how to chase and experts yeah what there is if.

D

That's not already done it's.

A

Already a writer so.

D

Yeah so I think then I think the I HD is fine Oh further, explaining that it's still a good thing. Yeah all the updates.

C

Have been submitted to the their authors, the authors of the original in eighteen comment, so we are just. We are waiting for Ben Campbell's feedback under discuss.

A

A

So I said this already when I started this session and there's some comments also from BJ who thinks everything is good with the documentary some minor comments sabine, you gonna address it and then we're gonna, kick it back to the is she and then hopefully this is another one. We got done all right so.

A

So yeah, actually richard just came to me and and I- think it's good suggestion. You suggest a minor change in the agenda. He would like to talk about path, vector and SSE first and about then about FCI, which I think is good, because those are more mature and we want to get those really done. So if nobody has any complaints, I mean we're not changing any items on the agenda. Just the order Richard has three presentations is going to start with M path, vector first and SNC.

A

First, okay, but I think nobody yet I mean you're not taking more time you're. Just changing your.

A

Ss is fine, let me just open it.

A

Yes, there, it is.

A

G

Ahead Richard, so let me get started and I'll talk about SSE and we all believe that SS e stands for server, sent events and is one of the very important capabilities of Otto, and it would be not only for any previous services and we want to be as young as possible so for this will be very, very Erica if you believe we think within you about Auto, to eat service for networker information updates, we discuss a lot about a possibility or whether, for example, you should be too given that which make it easy to would replace.

G

Ssc are not with so actually with people's who feel very strongly and SSE is a very, very interesting, very clean. It very easy to use interface. Therefore, we appeal strongly that we should finish this one and make it as foundation. Okay, so.

B

G

To about details- and the current version is 17 clearly has gone through quite a lot of changes and for portion 17 and for people who really want and can take a diff from the 30, which is last version and n-470. We did the submit capo.

G

Now the only two versions we did submit a capo in communities between those two IPS to give people a chance to really read a better version and the final version is 70 and there are two types of major types of changes being SSE on the first major type, of course, would be essentially is the issues identified in Alaska and we did God, allow the feedback and, from all the people, authors themselves from crab a few other people and so on and I think some updates are not really meaningless.

G

There were a lot of private emails, how the private discussions and so on, and so we try to fix our issues and, second out, of course, I said we didn't mean to issue which was bugging us for a long time, mostly the bad pass factor which I'll talk about next, and it is an issue about how to handle pass factor, acted more generically and how to handle our T's about multi-part, because right now and obviously from updates, essentially limited to a single JSON object and for being vision, I even show the allowable resource would become multi-party to be a motility sign.

G

So therefore, we finally decided why don't we just fix the issue in this version itself?

G

So therefore, that's why it took a little time. We believe that we finally solve it and never corner I. Don't want me to even beginning a huge group efforts by many in several product or quite a few weekly meetings with credible people with all countries and meetings and details, and so on. So now let me go to the details and about all the changes of we geek and I.

G

Think maybe it'll be hard for people to read, but I did a bundle show can a comparison between the older version and you're working with a version 17. An idea want to show the other changes that we have made, and here is outline of 30 and also that's the 70 on your right-hand side. You can see that a major change was this reorganization of the structure of the document and it does not involve major changes of the content but doesn't involve major changes of the structure, for example for people who carry in the old version.

G

We do have a lot of time discussions. For example, we talked about to our server and now we decide that's eventually giving at least stage it becomes less important to really talk about oddities and options. Therefore, the decision was to move into I, don't know. What's the protocol you've been trolled would be, would be mana happy if you even chill RFC editors, all yeah I said: okay, that's enough about decisions about debate you can remove. Therefore, we mutually moving into the last section.

G

Remove that section on what to define become this last section. We don't refer to anything in the earlier sections. The secondary media, restructure of a document is.

G

For example, at what time and you do, for example, for update what time you do incremental how you do a deal can tip in maps. I eventually realized. That's, essentially our operational and processing counselor. If it's so, therefore, if you look at it over here, which will organizes all these considerations into single place, and we view that that's a very well structure to change to organize everything in a single place.

G

So therefore there's one major structure, change and the second immediate structuring, of course, is the security and look at the over there and on the earth version 13, and we have all and but later realize, oh there's, backup with organized all secured attacks because that's its owner. We learned from all this su reviews sort of what happens. What now we have. Second, we give all kind of different types of security tax which I'll go to the teachers, a workout on a second part. So now let me go to the major types of changes.

G

Now they mean really highlight and la major components, and now we have defined esse and in the 0t final capabilities of SS and number one. Of course, this one will be interesting. Is it provide incremental encoding, I.

G

Can see over there and the second major change actually the moment we allow concurrent changes to be really pushed about multiple resources, and for that part, and you can look at it so we highlight over here- one metre chain is in older version.

G

Its upstream is, if you have one obvious stream, you said a lot South Stream. Therefore, now we remain in to call it this fast. You right, I think that's one very clean, very well-structured. Instead of Kline assigning us some streams. Therefore, I'm just remember part of what we try to clarify about all the meter changes.

G

I think that's been incredibly useful, a capability, and now we can really update multi-part, centrally single JSON object, and now we can opt in more JSON objects. You know who is very, very clean, simple chain without highlight and.

G

Neither was done before 13, but I do on highlight and we rename it like a control. Now we say: hey look some right where it.

G

Will not become very powerful and single, even if I had a control mechanism, so those are the summary of changes and the wording was also I'll get over there. So now, I'll. Of course, we define later beta about all the terms and I wouldn't want to go.

G

You know document because we did get a lot of feedback, so okay, some were a terms are defined in prod, okay, we are doing other a little bit because I thing is there, so that what we do that part and with you're pretty calm, happy about that, and now that's probably one of the Alameda change feedback from people, and so that's the system, diagram.

G

G

So those are the.

G

Choices, for example,.

G

I may be item from summary. Readers was why suppose I do have, for example, I have I, can still a base protocol, encoding and eventually assume you've been told. You know I lotion and you have five types of increment: do some hatch merkat. Maybe these are all the other types of essential, incremental encoding methods. I know why don't you a lot of it to indicate and neither can see it's a subset. Okay I can only do.

G

For example, the Jason do some more I cannot do any other incremental update their handshaking protocol and you see what they can offer delegate care, what they won't choose and so on. But you know, given decision still is.

G

By the clients and then want to meet the system slightly more complex and move you out of the auto eventual a person that could be server not on the client and the collateral, implement encoding matters as much as possible, and so on, so for more funny great handshake choices. So that's a major.

B

Design decision.

G

And we clarified it in this document and the last one.

G

G

Stream ID, but if you.

B

Only keep stream.

G

Id and for all the protocols that identifies with all of these resource, that's not why we have multi-part modular design and different objects, put into different objects.

G

Just fixing an.

G

Id on the pass, the object, IP and okay I'll come to an ID.

G

We think that's very beautiful design and.

G

You know everything will be fine, now handle everything. Okay. So now let me summarize.

G

So far- and we are.

G

G

Comments from today, if any any comments, please do tell us and if you have not read- or you know that you have not read only to the question- please give us and we want. We want to really run on you possible as possible, and we want essentially, for example, finalized for the final version. Reading Wow I know.

A

Work through all razzle.

G

Process right sorry guys so.

A

Thank you Richard for that presentation. I think this documents are really mature, I mean it's version. 16 and I just went through the.

B

Mailing list after.

A

The first working group last call you got extensive problems from to review us and make sense of other comments, so I think we as me personally as chair, think this really is ready now and regarding process and what we normally do. Is we issue a second working group last call and I. Think two weeks is the standard. We're gonna do that my question to Mia would be I and so I looked at the mailing list. I, don't think we need a second review individual review this time. Do you always wear?

A

The working plus call need a detailed review from a working group, member, no right so, but I would say we had to. We had to individual reviews with the first extensive comments, plus other comments from in the work, and you have to address all of them. So I would just to do a proper process. I would issue a second working group last call for two weeks, but no no individual not require or not request an individual review right. Yes,.

D

Right and so two things, if you want to start the working group last call immediately like during the ITF week, then it should be three weeks or you started after the ITF week. Then you go for two weeks in the end right.

A

D

But it's like they don't start it now, and only put it two weeks in because and.

B

D

Comment is it's very useful to you? Don't have to you everything, but if you last portion look at the changes and then send an email to the list saying everything is okay. I looked at it great, like you, don't have to provide a detailed review. Anything like this, but.

A

Yes, right right and in.

B

A

In any case, this has been a significant update, so we we have to do another working blast, call to give the working group a chance to comment, and so thanks. So we will issue a working group. I. Take a note, will issue a working class call on the email list after the meeting, but this seems to be ready. Then I don't expect any major things. So this is another.

B

A

Write are there any other comments No, and that means Richard. You can move up to path.

G

A

G

To excellent depends on this one, so let me get a started data.

H

One is slightly.

G

More interesting.

A

I'm just trying to find the slides, but maybe it's.

G

So Richard, please go ahead. Okay, let me talk about this. One. An occultist vector essentially, is our topology, we're kind of different from eye to eye RSR, the topology encoding and so on, but we don't want to really overlaps of work and for people who knew and the path vector essentially, is the topology encoding of a network. But we decided I thought. Oh, we don't really want a network element and type of the topology encoding at all, because they're too low-level.

G

We are not really talking about controlling every individual device at all and what our really wants is a very high level abstraction of a tautology. Therefore, that's really the AB structure. We're talking about and pass factor is that's why we think that's one of a major piece of work, and we also did a lot about this, so hopefully I've. We also feel that this version is mostly writing and but there's more dependency.

G

We got reason at the end and but I do want to hear people feedback, because we feel that this one also getting ready know. So now, let me just give you a really high-level overview about the water. We have changed in this unison, newer version, oh six and a major technical design chain.

G

That was the baking bomb actually back in the first, for like two or three times, I, never just go back and forth and go backwards. We give it all the time, but you've enjoyed this moment and all authors and I could not only author. A lot of people remember I talked about the weekly meetings with prophet and picking on the people every single week and most people through that we got it right. So we want to be the guys are working groups. Also, people see, do we get a break or not?

G

I never had a lot of like small text updates and clarify approaches.

G

The cosmos, because that's named you and truly- and we finalize it Anthony domain and proper the time registry by a following latest standard, install so therefore that's an overall structure. Well, the meaning part is I, want all the first part, which is the structure of the system. Buhl Adam would get most important part and we really won't get a working group accounts.

G

So here is for people who need to refresh your memory a little bit, and here is what this abstraction really does- is Auto a pass vector so fundamental. What character that does is the following is I have a lot for exam. This is really designing for the context of data analytics as a multi between use case, and you have a lot of data, for example, spark and MapReduce and all the other ones. Often time is a large number of potential, for example, my person with you, so you won't send the information around send their wine.

G

We sell wine sender to receiver to and so on. You have all the possibilities, and what do you want is get a shutterbug the resource capability from all these simultaneous flows? What do we call closed and that's you wanna say: okay, what I can count really have four if I use always flows together, what a country I have for the other one. It turns out. We the convenience kind of abstraction.

G

What are we you types of data type of one, but data is what we call it cost map it's simple, the past vector. Basically, we will say: okay and you wanna.

G

Go over those elements, whatsapp Network after element, it could be a single network switch, it could be a sequence which is, it could be whatever. Essentially this is there abstraction?

G

You might have multiple these closed and the open would be be. Chocolate actually weigh in at the same time, but if you only have a perspective on a bead, if you don't have any not in Commission, and one thing you do, is you also.

B

G

The private element, for example, say: hey I'm, using this one and this one you need to tell a user.

G

And you build a matrix and you could you can computer share bottleneck if you don't get properly so fundamental? That is abstraction with a repeating this very generic element.

G

Then we faced a major to that probably is protocol and me to dinner. If, in the same message, why did I option, one which was discussed in the very part, was very simple? I was that's.

G

A very clean design, but cookies one because of why using media.

G

Design and for example, you talk about llaman, you don't care about like a modular transformation, your honor one learn the type of type system as total, or didn't view that less too much. That would a brick or design you know what happens is.

G

So in the cosmic become watching to and other cavities that's another generic solution and we realize, like carpeting of resources, you follow me sure that really solve it nicely is that it.

G

Encoding everything beautiful get covered, reusability model, design and nice nice things you can see about that now. Let me go.

G

So what snapshot consistency and you do is follow the query you want to this abstraction I'm doing start.

G

G

G

G

That's very nice design is a working group in all three, but unfortunately, this one doesn't work. That's why the team's star for Wow. So what are the issues now? Let me really go over issue number. One about this design is.

G

What can massive you get back from server for service for this URL.

G

It's a cost map, this network map, and this is a pass vector, and this is an endpoint or cosmic. You have all cancers also indicate from it.

G

But now, let's do a very simple thought: experiment, abyzou, okay, let's be consistent and get back from sir.

G

Okey, which one is that we, the Falak, awesome how you can improve on it: that's not elegant and world war stuff. Oh, we really missing.

G

And were stuck on this issue, the second issue we also stuck- and also we work on to really handle I- think very carefully was I talked about SSE issue.

G

Now how they use SSE a pre-designed myiasis, it was very simple: you have the object indicator. Essentially the stops communicators, but here is.

G

And I don't know.

G

G

Crimes beautiful.

G

B

G

The elements, okay, enough content, let's really illustrate and don't want to read it like it- is like a tax mess text and let me show by it, goes through a whole process. A few messages on what's going on, I think this we find like you. Rather we get everything correct here.

G

Okay, so focus IRB based.

G

On past vector from phone about some any pop in put in a host IP addresses to all the other ones. I wanna use pass vector to get all these abstractions. So what I specify.

G

Now you'll be nice, oh especially for all these essential imports. All IP addresses.

G

If you won't do another query and you want to.

G

Know by looking as your Pokemon, because.

G

Y-You just have exhilarated centrally dependent of our European path.

G

You know what service it is I now with IRD I. Now the clan issue occurred. Okay, so here.

G

So that's an example where you query by input so.

G

Karaoke game and then what you do is is specified all in pumps. You want to be the credit for here. You have two queries. You're gonna send simultaneously and you guys I know what happens is response.

G

Let me just go over for people.

G

Now, here everything can also work very well.

G

Everything just works so now, let me go suppose: noise I just want to do incremental updates. Okay, so now, let's see how you do increment updates is also not turning out to be very simple.

G

Extractions, you also, and this one only this one would try increment update. So what happens.

G

With abstraction, okay, so, therefore, and what piece that you do is follow, is you go and you will add a subscription service? Oh I, see so you do essentially is especially buy a new subscription service, for example. Here it's easy as pesticide subscription, one it's very clean, and then you get a data and then whatever is fun, look like it is event. We've indicated what type of you get and then what he also specify.

G

And then you've got a.

B

Comment because.

G

You really for often time for increment updates for these kind of push notification. First,.

G

Simple and here's the first message and have a whole message: I know what you do is and that's the subscription service, the company I'm now the two pieces to part individually. You can say: okay.

G

He which this object was continuity. I, now add these two objects individually, so everything now I'll just work together, really nicely I, not only summarized discussion but also sphere that, finally, they got it right. It took us really some time and.

G

Feedback and we're gonna wait for SSE and because we depend on our SSE and unified property on my property and then we can proceed with finalization. So those would be optimum pass back to your document.

A

Right so, and thank you Richard for the explanation of all the work you've done. I know that there are lots of rather complicated issues that you addressed, and so my first question would be: are there any technical questions and then I want to discuss the process issue.

A

There seems to be no technical question so regarding process, yeah Richard I mean it's. What is your view? I think this is ready for working class call. It is okay, so, as chair I, don't have any objection, I would clarify with Vijay, of course, but I. Think if you think you have solved all the dependencies and you have updated you've informed the working now, everybody and I think I would discuss with Vijay but most likely because on path vector and you are going to talk about unified properties. Yes, I think.

B

With next property.

A

Right right so, but yeah so on per factor, I think I'm. Gonna. Take a note. The Vijay.

F

But we're most likely.

A

Gonna issue a working class call and we.

G

Are more than happy.

A

G

Finally, and of course the only a caveat would be a, for example, there's any last name of the change property. We need a pillows. I presume also be examples. We don't right.

A

So maybe I should have set this also in the beginning, a while ago we agreed that path, vector and unified properties. I want to be moved to.

B

A

Together, so obviously that's the caveat we have so if there is any issue between the five properties I'm, not so sure about SSE, because that one is mostly done but yeah right great any.

G

Other comments.

A

No great okay, so are you you are going to present now on FC FC is, if you, if I, probably sorry, yes, let's see if I can find.

A

Yeah, so who's gonna present Sabine's got present all right. Sorry, I didn't right.

C

It's actually, we are going to share the presentation with each sub. Well, on that one we had a say yeah as well substantial discussions, but because it's it's very important I mean it's really central to to perfect on to the current chart. Actually, so.

C

Basically, so, since the last version, we were trying to figure out how to couple our how to build a consistency on a property and also the the resources that information resources that are used to convey this property or to convey the entity domain. The of the objects on which these properties are conveyed. So this presentation is halfway between the current version and the one that will be submitted after this meeting so the well. The current version is, oh seven, so, basically the updates. Well, there have been major changes in section 2, I Section.

C

Do because that's where all the definitions are. So we visited definitions of unified property where a property type property name dependency type, and actually we are not done so, but so I will not focus on what we did, because we intend to change it, but I will try to focus on explaining the problem so well.

C

One change that we will keep is to move from the notion of entity address to entity identifier because well address makes sense in ipv4 and ipv6, but in like in, but in the PID, for example, that it's a name in in country code and what-have-you. It's well, it's hard to eat better to have identified. This is more generic, so we proposed in the current version, an option one to solve the following problem. Let me start with an example. Ok, this example is inspired from the FCI LCD and I draft, and in that CDN I draft.

C

There is an example of an eye IRT and actually look looking at the area D. We figure that and looking at the example, transactions requests and responses. Well, we figure that we have a sorry, so we have a server who would like to convey properties, say: CDN IFC, a capabilities and property P ID, and it would like to convey it for entity in the domains, ipv4 ipv6 country card and ASM. So that's the example.

C

So, basically the server says: oh ok, yeah, it's kind of convenient or I often have request in that sense. So let me put a map property map where I convey like city, ni, f, CI capabilities and PIDs on those doubles and may be given some requests. 7 would like to convey all these properties, but the problem with when you, when you have to document what proper, what map is used to verify those properties, it's hard to figure out which one relates to it.

C

So this is actually a bad example here, because it does not allow to to provide the resources dependency without introducing ambiguities. So, okay, there is one one way to figure out how to do. This is okay, I can say. Well, ok, let me let me have a property map that conveys if capabilities and PIDs on entities in this domain, because it's a frequent request say so that relies on my default network map and my default CDN IFC, icon and I can have another property map that defines CDN IFC capabilities on these entities in this domain.

C

So this way I do not introduce and well introduce less ambiguity and I avoid associating entity domains, resources in entity domains on which I cannot define the property. So I do not introduce arrows, but if you look at that, you can say well, it's just redundancy and it's still not clean. There is no clean, an ambiguous way to to associate resource dependency with property map resources. So so now let me go back to a couple of slides so and we have these characteristics on properties that some properties can only apply to given entity domain.

C

So that's fine. So, for example, you can define the PID property on ipv4 domain, but you cannot define it on pids and another point is that the the the the the understanding of the value of a property depends on which domain you apply it, for example, suppose I define a property called geolocation, so that is encoded as a couple of latitude longitude.

C

So if I apply to a entity that has an IP for v4 address, which is just a host fine, okay, that defines the coordinates where this house is located. But I may want to define it to a twelve to associate two so a PID okay, then I could define it as the coordinates of the center of gravity of all these holes or of this domain of this sub Network, and so in that case it would have a different meaning and in the end first of all, it would have a different meaning a second.

C

Then it would depend on a given information reasons. So this is just to illustrate the real necessity of having a mechanisms that really in a clean way, associate resources, dependency and ambiguous resources, dependency to on property map resources, Sabine.

A

And we are slightly behind schedule and.

C

A

Saw you have still a lot of slides.

C

So I'll give the flow to Richard.

A

All right I know you have material and I think people still useful, because people can look at it, but.

G

So here is one way about centrality is an option, one, that's what we write up in the karma craft and for people here is and their own is rules, and so on, basically, I think its objectives already overview and so on, and the major you get out from one is binary sort dependencies to the whole property map server. We feel that eventually and to handle all kinds of very flexible design we Amalia Condor, is no longer as clean as possible.

G

Well I'm gonna talk about is design option two and well. I know how many people here are experts in a relational database, an expert.

G

A

Assume people in the room know what a relational database.

G

Great great, so we do it have the right audience know we can kind of huge amount of the issues to really clean clean and so eventually, actually what I propose by the way, even though it's not like a meter check is not mostly, we all 2.5 2.6, they are not meter kick, but somehow, unfortunately, they will give you a pull of different view about world. Look at how we solve this from circle diamond to get started on option to even design resources. The query become unified to become too complex report one option to so.

G

Therefore, let me just get started.

G

Which is you look at every single I will with all the base protocol. You can consider this one as just set up a database tables I. Don't if you get this, that's very clean and so basically.

B

G

A network map it's just called here is our visa object, an aggregation of a network nodes and what we represent is very simple. Is you write it down as network and there's a table additional people name.

G

Tickets and name prefix a name prefix into it. That's a beaut! You look at it. You have a nice few different part. You do a different aggregation.

G

Okay I now, if you move how you're doing in from a query essential that you're doing a sequel select, so what happens is.

G

What IP is our this essentially can be accomplished a very simple cycle curve, so for people with me so far, that's a dramatic point of view and, of course, this algebra.

G

But the work can be handled by something.

G

Comes tables tables, that's what we do I know. If you talk about this, why not I said I want to propose different? You I? Could the qi- you very much traffic for each resource in other define a table in a database? The whole database Alto server, is a database server and maintains all kind of tables. Of course, with different semantics of you have complexes. For example, proof is my chance on a column of table, defines an attribute.

G

The projection of column define is to meet those all standards, sequel and relational database definitions and, for example,.

G

Protection every single column, you get a domain.

G

You table defines of automatic, for example, people want to learn a few from my people, you for working pit, I'm ready, you can even build a river smack. You got you're gonna database, a single database Aggie. What I see a.

G

Problematic we find earlier from an attribute to a property. Oh, it does therefore univer pervasive it's a sequel query so, for example, for the crit example for this one very quickly and what is it Sucre? What really? If you really want to write down this one has issues using, even if not greater purity you're, my very favorite numbers with overlapping addresses, but conceptually essentially is customary version of circle. Query select, ipv4, never connect one from I don't want to. We are no input you suck all about it.

G

Is it's beautiful to see if you can handle all over like incorrectly lovely reviews, especially so, if we talk about this one now they sent design space for not like me to design each piece of what do is given number one, but you can enter this, but if I enter taste, we can go to contact, call and name space. I can for the column, is an advocate of a name of to you. Every database I knew you create a table, five columns, each column, you give a type and name, and here it compiled.

G

So what it is int in this string and hear me was it domain is very simple. It's that indicate the resource. You do a projection.

G

And I think you know maybe come there somehow.

G

Which people Richard, okay and then in detail. Of course, it don't IBM specific and probably become name and go or probably identify a property domain and.

B

Therefore, essential.

G

So now I'll wrap up so basically now what it will happen there simple is is very simple: it's also resource you define you define a set of entities where an entity is domain and properties for that. For the meantime, you just essentially.

G

As a standard, and if you do this I.

G

Owe my resource its own property and so on, and probably you can flow this way. I could can be recursive process by itself. So if I were to only difference in your problem at all the resource resource, ID simple, if you can use it this, your sub pieces to locate this myself, I've missed profit myself, otherwise, import from some resources. You define a total of you so therefore, and basic design and what types type system we handle okay. So at least we have, we define him and working for the map for every we.

G

Never you divine you also cell type service, for example, and then.

G

You find this instruction I know what order so all the dependents they have all these complex audios in the organ we just define a very clean based on the relational algebra algebra and everything just go away and I'll just come. Let's go away, we have very simple, clean design, so that's my and I think for people waste database background. This rubbish should look very, very intuitive, so discussion will be, which 1000 years and, of course, what we will do is we'll post all of meaningless.

G

To begin to talk all the designs, people monologue come to come discuss way. One thing little new word graph for the copy cut. I think this change actually is my new teachers or properties and refuges. They are not immediate exchanges and it goes.

G

A

Thank you, Sabina Richard, so I think for most of the working with members. They really want to follow this. They need to read the draft and as follows: I mean you gave a high level explanation, but it was probably quick for people and so people who are interested in this. Please read the document and give feedback. That's my comment regarding process and as far as I understood, both design options are not any more the dependency you have perfect right, so it does.

G

A

So let me rephrase my question is: can we issue a working group last call and path vector, while this is still undecided or do we need to wait? That's what I you I mean. We can, if you say, Pat vector, is more or less done, but we should wait until this is done. That's not a problem. I guess and one.

G

Way we can do, is we can I be if we probably can remove a dependency of unified property away from to this, there.

A

Will be an option and again I, don't have any strong feelings if we want to do it right right so I also we can wait so I guess you should discuss with the authors of both documents, what you proposed and for now I'm taking you note not to issue working class now, and we will have a discussion on the mailing list and see how it goes. My understanding is the path vector is more or less done on mas, yes and you're. Just waiting to resolve this issue for the result from Eagle, Pass.

A

Okay, so then I know I think about it. I think we waited it, but all of a Spectre just to make everything. Okay, you know properly, and so we hope that you can resolve this. I encourage you to to ask people on the mailing list or people who are interested in this comment. Sorry again, I think you gave a good overview, but it was probably nobody.

B

A

You haven't read the document. It was probably too quick to follow everything and yeah. So that's my recommendation and then obviously as chairs, we want to get it done, and so please do, as you promised, discuss your options and sure. So as far as I understand you won't once you decided for one of the two options. The draft is ready for work. Yes, yes, so.

H

Yes, just 160 170 mm at.

A

End of April I think that's good yeah, let's aim of vibe and and and then also with the path Factory. We can wait another couple of weeks cracking we decide, Pat factors ready, we put it on hold. You decide on one or two options on the mailing list, and and among the authors of course, and you make a decision and then we can issue working desk. All combined for both documents once you've updated discipline sounds good. Okay, okay, any comments, questions No.

A

So then we are five minutes behind schedule. Richard, you are up with FCI, yes, I, which.

G

A

So I mean the agenda has been set up, that the working group items have priority so yeah. But if you can hurry up be, we would have time for the three individual presentations and I think they would be thankful if they have their so yeah I'm. Putting up FCI I'll.

G

A

Let's see were actually.

G

I feel motive is a lot of designing for this unit. We don't really unify everything, sometimes if you only really ambitious. Here's FCI all right. Okay, I'll, be brief, and the quick so basically I've, seen stock on the relative, stable and the encoding. So basically is it's quite a stable, of course. Now the meter from, of course it depends on SSE and in Section three point: seven, three, three six point: two four different examples: we really use examples, only-- and unified property in Section six and an eight we depend on using properly into this example.

G

So those are the dependencies of FC about. There are just small, small, Nuggets or very small examples, and the changes are betting. All we've got a lot of reviews and the trainer addressable video six, and there are some issues raised by reviews for the addressee. Aloof I'll, be focus very quickly, I'm, just keeping all these reviews again.

G

And so on and I travel can go over on the definitions, regions and secure company. A lot of significance, I saw I, probably can just keep and they are not really essential. They're two major issue and Kotick they're not like made a personal I, thought some reviewers. Some review very strongly.

G

And issue number one from one of the reviews is the following, but right now we call FCI footprints again: I footprint, a footprint and capability interface week, however uncut in math and in review or my review were basically does a few likely feel like at all. And here is the the argument from our base protocol. This document also defines dictionary maps or maps.

G

For short, this means you know the base protocol Maps defined as a dictionary maps, which is maps from streams to change some values and, for example, bucket below P files type three: are they, as might have one math? We define a string and you must define any type.

G

Basically, okay, Type three Mac on five one, five, two I'm sitting I now the argument would be we're using the definition from sitting a working group and turns City and working with decided now that you least got a key value store and in secondly, use, for example, a different format. Pc me: okay, okay, baby, the type and the capability that type value and a further print I'm coming about it, not even key value store and therefore the objection from that review is.

G

We should not do the name called map and revision is very simple with him like a maps and keep up with him, we finally table or entry or data, and so therefore- and personally I am you know, I'm very new to. But if we wonder strong feeling from working group all we can change this name. We don't call CDA IFC a macro Ksenia FCI data, whatever table, because we really using a normal we're done is the key by the store, which is typically like reading a lot of public databases. They, because there's also aa databases.

G

Only the store, I'm very neutral on this. So therefore that scripture number one the current design allow square only by footprint or only by capability and reviewer is to talk about. Can you do the query for both footprint and capability? And oh, we have two fairies already and number two document to be a person and.

G

First of all, I, probably a couple ways up to number one, but I do want to pose the issue to the working group because that's the minimum there, but we can finish document as soon as possible, but I do one bottle. They are the issues I think in those are issues all the rest which attack or it down, and therefore I'm looking for these two issues. So, let's see if any suggestions or feedback from people on these two issues. If you want you to thank.

A

You very much for being quick richest and have for people interested in this document. Please send your comments on the mailing list, so the authors know the opinion of the working group. After all, this is a working group item and if there are no or few comments, Richard I think the authors.

A

Because, regardless.

G

H

A

And I'm happy that this gets.

A

Thank you so that closes all our working group items I'm happy that we're making progress and we decided already that to working so basically cost calendar we sent to is G once something gives an update and SSE we're going to do a workgroup last call, and hopefully, after the one issue on unified properties, have has been resolved. We can also move path, vector and unified properties to work on last call and CD and IFC is the one that maybe takes one one step longer, but also that we can hopefully move along great.

A

So that means we now have if I'm correct, 25 more minutes 610 the session ends right, anybody, yeah, so 25 more minutes. We have three more presentations. That means eight minutes each and the first one is Danny I. Think right: okay, please!

A

So these are individual documents that we are happy, that people are working on them and we try to give them a time to present, but, as I said before, we're not taking any house on working group adoption at this point.

B

In time, because.

A

We want to spend our energy on finishing the working items or at our milestones. So this is you right? Yes,.

A

H

Is Annie from the University of Campinas? Basically, the rationale of this. This document is to offer an overview of externalization activities and research projects, including the problem statement behind the deprivation on end-to-end network services across different domains. In addition from a layered network model, it's also profit or opposite on potential Altos tension related to n to a network service requirement representation, and another important object is to to twist our discussions into the alto working group on the industrial players related to to potential new eating's or use case.

H

That could motivate the need in the new alto charter was the the current working group. Eating's has been completed, basically entering that or services often required if the network function and there a specific order and network operator are considering to affect the practical customer, the possibility of configuring network services with the specific requirements in terror of resources or perform at objectives. Such demands are usually composed by distributed resources which are twist expected available across different domains with different technologies and administrations.

H

Int. If contact difference, if standardization activities and research project as been focusing on multi domain intended or services, for example. In case of the alto working group, we have the the some documents.

H

For example, the alto basic block addresses single protector, provides a broker plane working as a coordinator between a set of and top-level control planes for multi domain orchestration and the unique or document provide a resource, orchestration framework for multi-dimensional distributed that take analytics and another document that she will present later introduced several multi domain application and how they can be benefit from from from Al for any Alto protocol. Have another initiative between the survey function, chaining working group and between the XE, for example.

H

We have to report that that provide difference and a Freeman architectural approach with use cases related to network services across different administrative domains. In case of the research priority, they included, kicked actual models integrating and it's beyond lding to order the challenges to our flexible multi domain service chaining. We have different approach. For example, Beatle antonova provide a centralized approach or or in case of D, Phi, DX or fudgie transform it and a forward they provide a distributed approach and related to to the to.

H

The problem is Taming, as I mentioned in before I mean to in service requests. Describe it well, not a set of filter. Network function are built on links that determine the order in which the built another function need to be executed. Both little notes on it elite are links, are limited resources and can be local in different domains. So the deployment decision problem involved the discovery of bets candidates, domains or the source embeds, feasible paths between such results.

H

The in order to the toe to take optimizing decision for further placement, the managing application or or NATO application need to maintain a never inventor. In order to to get a real-time representation of the available infrastructure resources. The size of this network inventory can be very latching as energy such as distributed cloud or edge computing, so aggregation mechanisms are necessary in order to reduce the type to discover such the source is all for anime for the mana dreaming of them.

H

Once the network mentor is built, a publishing information mekinese is also necessary in order to the elementary can provide a simplified. Yet enough information bill to the network application and just to see quickly the exe reference architecture, but by far the most the most respected reference in every reference architecture. Here we have here we have the only lefty that they playing on. We have the NFB infrastructure, we have been FM BNF, an element management system and on the on the right, we have the control plane.

H

We have the vehicle infrastructure manager, we have the ENFP manager, PMF manager, and if we orchestrated on the top front, is in a free reference architecture, we have a layered.

B

And network model.

H

At this model, a lower separation between the network elationship in different levels of abstraction, for example. Here we have on a twin network service where composite by a set of different eigen functions represented by the network service layer. Each built on either function is composed by a set of beautiful functional components represented by the ebony flayer, and the reservation represent the virtual physical resources where the vehicle function component will be instantiated so interrupts, depending of any free architecture that that we'd use in multi-dimensionally.

H

We can have, for example, an f-18 for us to us a service as an alt information sources, and we can come in onion if we orchestrate around or BNF manager as an alto clay looking for for domain. But if than say the vnf or the butyl function, components from front this layer in the world model, we propose a an attestation related to n n networks. Every requirement.

H

A presentation on initial proposal is presenting in this draft deactivated broken a system with America Sater, where an Edward application working as an alto client, need to specify the set of requirements to the alto servitude tank and the data source or domains and candidate. But the idea is to use the X in F EB mono data model that provide that defines the network service is a composition on the door for sure, including the functional and behavioral specification. It's such a specification or cut rate in templates, collect network service, a scraper or BNF descriptor.

H

Yeah, basically, the network service descriptor, it contains static information, use it to the processor or from boarding network services and the network function. Descriptor provides information, use it to the process of boarding buta at your function. I have finally a snip step Sidious to to get that feedback from the from the alto working group. As I'm mentioning when I started my presentation, the idea is to to interact with the industrial player in order to identify news cases or that could be beautiful, justify the retarted or possible new station to the alto based protocol.

H

In the particular case of the university of campinas, we we have a couple of interested in this airplane and we have telefónica ericsson and we are trying to identify specific action points related to this extension. That's it. Thank you.

A

All right, Thank, You, Denny, you're, absolutely exactly on time. I would refer questions to the mailing list so that we can have two more presentations. I think that's in everybody's interest. Certainly if we are rechart on the chartering, the working group, this work that needs to be considered, then I totally agree, but, as I said before, and we, if we're gonna, reach harder, we also need to find new chairs for this work.

A

So, let's yeah. Let's keep this in mind and thank you for a presentation and if there are any questions, please raise your hand on the mailing list and now we're gonna have two more presentations. Thank you very much.

H

A

You so I'm just talking at the agenda and who's next, so Jensen is next on multi-part Alto I hope that he's gonna present remotely I hope that that's gonna work again and meet Eko yeah I mean with the the meteor go team, send him emails and I think but I don't see him online. So oh yeah there he is Arius right.

A

Everything works great. We can see you okay,.

I

A

Yeah I guess I can move the sides forward and you can talk. Let me just open your slides. Give me one second, and this is multi-part right. So I have your sides here and you can just go ahead. So this is now slide number one yeah.

I

Number two yeah: okay, thank you, sue heroin is Stinson and we keep some athlete on the out of multiple extensions day. So this is already the third revision of the automatic document and god this revision is want to clarify some motivations and the some repetitive compatibility shoes, the mango and since the last revision from the last Auto interim meeting. Ladies following changes, they write some term definitions to make consistent with some related RC and follow the reps on education and examples to follow changes.

I

Often relative working group documents like the SSE and the unified property, and also make some major change to extend our to that space and saw some credibility, compatibility issues yeah next page, please, okay! So that's a that. Just pays the new trance base for the kind of document for this turned out of multi-part extension, and we spiced indicate three kitchens week. We can use the multi-part messages in the auto protocol, so the first case consider it called multi-part request and the multi for response and in the or a single example indicated bath.

I

Curie, the crank has sent a mouth. Multi-Part makes the request, including a sequence of the auto request, and the server returns on in also in a multi-part makes to respond yeah other and the second case. We call it a single request and the fixed layout multi-part based bound and a specter Curie about a mentioned in our previous our previous session. This is a special case always and in this case, meteor type of the request specify a single test object, input, but the response it's a multi per message.

I

So, however, the layout of the response must be fake, so we need to expect eight expert, what's a layout of the original, so so that the client can predict what they response looks like and for the circus is similar to the second case, but a smart general, and we consider the response layout can be flexible and it can be change by the different requests, and this is, we already did cut and was in our previous revision.

I

So we do not waste time to about it. So nasty.

I

Yeah, thank you. So they see the example for the first case and in this case the in this example, the Delta server provider, IRT and export several resources, filtered necrometal and one filter cost map and option those of them used it the same before network map, so their services at the table and the felt or customer use the also use data for network map.

I

So both of the results used to sing the per network map and especially if I may define a multi-part resource and allow the clan to cure a post option in a single query and it can enforce a desire consistency and we saw without return the response in also in a single curry, also using single message.

I

So next page, please so for the second case, which is use the same example showing the phase vector document. But the IRT is a little different, because the pass factor is just a special case, but for the some common case went to mention that the the media type, the type parameter in the major type. It just indicates a the root part, and the krebons issue include the capabilities for the root part like in this case. It just need to include the concept name pad for the response.

I

The exact ramp in Manado was a response reply and so issue it completely. If part. What's the layout of the map part response message so in this example, the message should be includes a two part in the multiple response message. The first was a group based on coal, mine and just I can party to perform I and for a nice page, please yeah.

I

So in the our previous relation, we are addicted a except epic compatibility issues of some of the multi-part extensions, and only remaining usually how to us about the incremental update extension, but what early daily latest animated document already Chilean to mention how to install it and for the most part first case the actor exported unnecessary, because the current access, the document already allows the clan to to fight multiple recently in the same request.

I

But for the case to and history we call a stimulant precisely, which is already filled by the the past vectors and gently current accumulated extension. So when the clan subscribe, a melody Paris results at the first page update message: the server returns a full multi-part message and then the subscribe to every sub tree in this multicart tree, recursively and others I did a server, can publish the data updates for each node in this multi part 3 individually.

I

So let's of precise Rick post and it's more like the guideline around, but maybe the server client customer out loud.

I

We consider it can be specified in a credibility applicant.

A

And maybe one or two minutes, because we have one more presentation and we're running towards the end, can you come? Can you join us.

I

Already the last page: yes, an aspect where they're finding some real examples which can benefit from the nose for extensions, so one potential example may be. We consider is a multi democracy which other tahvalli's this topic and about DJ progress in the next village. So we hope they remembered can review and give some comments. Yes, thank you.

A

All right, thank you very much. I think the same goes for the Astros last presentation. If you have any comments, please do post them and if you want to review this document, you are more than happy to welcome to do so. But please do that on the mailing list. We want to have time for one more presentation, so Jensen Barry. Thank you very much. For most part, we continue quite well and see you. So this works quite well and I. Don't know! Where are you right now?

A

Oops you're gone already all right so ours next at presenting on multiple use cases, and so you see.

E

A

So I hope that all right, so I'm gonna just try to get up your slides. Give me just one second sure.

A

No too many slides this one right: okay, so I'm already moving to the second side, because the first one is just better yeah. So please go ahead. Yeah.

J

Hello, everyone- this is Chow, so the goal of this presentation or the corresponding draft is to give a review of important emerging multi-domain applications and the showed on how they can benefit from the out of protocol and after that, will describe the a generic framework for these applications to use Auto to include their performance and then we'll discuss a new design requirement for our two fighters for these applications. Next piece.

J

So this is the goal of the Auto based protocol. Basically, a Auto based protocol provide network information with the goal of modifying network resource consumption patterns, while maintaining or improving application performance, and the information neatly exposes as an abstract map of network, and the typical use case is used in the base protocol is the p2p pure selection. Next at least.

J

But with with new architectures as the emerging like and the new like, Sara started, changing and sovereignty for networking and all these architecture and techniques are deployed in multi domain networks, as such new applications would emerge when and these new applications and also benefit from autos. So here is the incomplete list.

J

So Danny has already talked about multi domain end-to-end network service deployment and in the past few ideas we have been summarizing our experience on developing a multi domain, clever data science, orchestration system using Auto, and also we are also working on if a new application called flexible inter domain routing control. So in the next few, slides I'm going to focus turn to quick, quick review, the second and third applications. Next, please.

J

So for multi domain clappers data scientists are different organizations with the country built at different resources to collaboratively plaid shirt and under analyze extremely large amounts of data. So the targeting example scenarios that we have been working on in the CMS experiment in the Large Hadron Collider, and we have designed a resource, discover and orchestration system called Mercator to Aqua, stir orchestrate large-scale data set transfer between difference and cmss around the whole world, and in this system we use auto cos map and easier as a service with pass vector extensions to perform resource discovery.

J

We took resource discovery. We also, we also used auto sse sir intentions and also cost calendar tweak to push the increment update and also to to for for future scheduling of the datasets transfers. We have given a demonstration as super 18 supercomputing 18. Last year I balanced to orchestrate the transmission of a set of science network flows from dallas palace pasadena at 100 feet per second, and our experience has been document has been documenting in previous drafts, which we will also plan to update shortly.

J

In the view, after this meeting next, please, and also we have been working, go back one more, maybe a two more please.

A

I'm trying to move ahead because we just have two more minutes so I wonder how many sides have I think it's too many.

J

Can you go to slide number number? Ten sorry.

B

Which one is that.

J

Number, you haven't saw that number eleven. Yes, all right, so yeah! So.

B

J

There are there are: there are different, much domain applications that can benefit from auto, but we also got these applications also oppose new design requirements for out of the battles of bodom. So here are also a incomplete list.

J

It includes how to expose accurate information while preserving natural privacy and how to expose the information of alternative resources so that the application can have more flexibility, choosing picking different resources and also giving networks with additional business business opportunities, and also you involve support, providing interface, more generic interfaces for more flexible curious from application and also provide a unified, accurate representation of different types of resources to applications next slide. Please.

A

Yeah I think we're running out of time, so I'm gonna go to the summary. Sorry well sure yeah.

J

We're running out of time sure okay go directly go back to the final slide, maybe right so I did just paste, basically just to quickly summarize that we have a tightly identified, the different applications I'll come back and that I can benefit from Auto and also new design requirements emerge and within the working group they're also having different graphs, both working group documents and personal jobs being proposed to address some of the design requirements.

J

So the next step we are going to do is to get feedback from the working group members and the industry partners and then we'll see smacking investigate investigate that is that to address all the design requirements and also the deployment concern of auto for these applications, including a stability incentive, scalability and previous and any others, and but here I hope. Yes, I finish thanks. Yeah.

A

I'm sorry I have to close now, but um the same work for all the individual presentations. If you have any comments, please comment on the mailing list get in contact with the authors.

A

Yes, so this closes our working group session today, I'm happy that we are making progress, I hope by mantra. We have further progress and that we have certainty. I'm, pretty sure we're gonna have two more documents out of the working group and maybe also to further ones. Working class calls finished at least that's our goal, so we're getting there and actually at this point in time, is actually realistic that we are getting done with our milestones by Montrell or shortly after which is would be really great.

A

Yes, so this causes our session. Thank you, everybody and they note note takers. Please send me an email with your notes or me, and Vijay and yeah see you in Montreal I'm, not sure. If I can be there, I'll try but Vijay said he's hopeful that he can be there and share the session all right. Thank you very much enjoy the rest of the week in the evening and life.