Internet Engineering Task Force 104, 25 Mar 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF104-EXTRA-20190325-1350

Description

EXTRA meeting session at IETF104
2019/03/25 1350

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

A

All right everybody, according to the clock, it is now 50 minutes past the hour. Welcome to extra session number one of two for those of you who who haven't caught up on what was happening with that. We've split this session, this two-hour session in half and given jam up the second hour so that people who are in the u.s. East Coast who need to attend can attend that session. We'll be doing the second half of extra tomorrow morning at 9:00 a.m.

A

in the session that used to belong to Gemma, and we will be going through imap4 over to during tomorrow's session today's session. We will be working on these things anyway. Let's start here, we have the know. Well, if you have not seen it yet, it's worth your while reading through that, so that you understand what your responsibilities are within this group we have had in the last season.

A

We had someone show up on the mailing list, with a set of requests that led us to eventually passing over to the Ombuds time and letting them know that this was happening because they were making demands and complaining about behavior. Other members on the mailing list- I'm hoping all that is resolved, but to keep in mind that this applies to your interactions. The ITF entirely.

A

Alright. Moving on to the agenda for today welcome Barry osser new Area Director for.

B

A

B

A

Looks brand new.

C

A

Alexia is still floating around writing stuff and involved we're going to look at the quota, the submission token and anything else, and possibly also talk about the rich charter. Today. Did you two want to switch orders? What do you want to still go with quota? First.

A

So, let's start with submission tokens in Mike there. Yes.

D

Welcome our remote people as well and yeah finally, Michael sloughs, ours and I have my co-author, actually the one who came up with most of this stuff Timo. He should be on that daddy gonna list, so hopefully Timo can.

A

D

One's ever accused me of being quiet, so hopefully you can can hear me in the back. There. Submission tokens didn't mean to kind of spring this on people too much. If you were at FOSDEM, bron-bron happened to join us at fives David, so he saw some of these. Oh, we.

E

Already have sorry I just prefer preface this. This is a not a working group document and at the moment this cannot be in the Charter no matter what so. This is the guest presentation present.

A

A just to say it's a.

E

A

E

A

A request that we yes and.

E

Then we can separately figure out what to do with it. Yes, yeah sorry.

D

That out of the way, dy I think the next slide don't have too many slides here, but this is really getting me into why this exists.

D

Actually, I'm gonna start with the third bullet point. First, the big project. One of the big projects we are working on is an idea of chat over email, essentially its chat over IMAP. It's not my favorite name, but that's what we decided to go with. Let's chat over email and part of that is looking at email and current email. Delivery can be slow at best, I mean you know, a certain 95th, percentile email delivering submission is you know it comes in good enough time, but these days modern messaging technologies.

D

Is you get this low, very low, latency delivery, with also things like potentially like confirmed delivery I mean. Obviously, when you send message, Justin, T, P, Q gone, you don't know, what's gonna happen, so the idea being that email, if, if possible, or an email system, if possible, should provide a more real-time messaging experience. So that's the problem that we are looking at and trying to figure out ways to make this better.

A

It's probably worth just giving it a brief couple of minutes, background, roar and what Caterpie map is and how it what it does.

D

Jennifer I mapped the idea being again: I I'm, not a big fan, I, see Vittorio's in the back here, I'm, not a big fan of the work I map on there, because it makes it seem like there's something special about I map. It really is more chat over email, which is we are currently working on and I.

D

Think I saw in your slide, I'm, not sure we're at the point where we want to discuss about standardizing COI or the chat over IMAP in general, because it still really is a work in progress, and it's really what this, what it's designed to do, there's a couple steps. The first step is to make a chat protocol or a chat work on existing IMAP or existing email servers was just essentially as IMAP. It's not gonna work we're not looking at pop, but we're looking at IMAP server.

D

So the existing pool of I'm f servers that exists. How do we make a chat experience that we can? We can make we can roll out today with existing technologies? You don't need to upgrade or anything, and then on top of that is looking down.

D

The next step would be: how do we improve technologies or either take advantage of extensions that already exist open extensions to make the chat a better experience, and/or, suggesting improvements to submission or other parts of the mail platform that would allow the experience to be a better experience against the IMAP part is in there because we feel right now. You know that is the mail.

D

That's the landscape! That's out there, a J map- we're not discounting. We talked about this before not discounting J map is something is, can be used in the future, but IMAP is just what what would exist in the you know in the wild today.

D

F

Is very so you're talking about having let's say, a chat, client that is using email, IMAP and SMTP and whatever on the back end, but it's a chat, client correct for my phone. Okay, yes,.

D

The idea being is especially the way that we look at it is you know, there's things you have a question we were facing at a place like Las them is like. Why not jabber? Why not one of these other technologies and the idea being is we're not saying you know, jabbers incompatible. Jabber is great. If you have jabber, has all these cool features, but the problem is, is most people in the world? Don't have a jabber account? Almost everybody. That's on the Internet has an email account already.

D

So the idea that you already have a chat account because you have an email address like that's it. That is your chat. Account is your.

A

Email, the other thing that I guess you didn't cover here that you did cover in the longer presentation, was that this is designed to fall back to regular emails and it's quite readable and usable just over email. But if you have this chat client, then you get things like it just displaying the individual responses, rather than the whole thread and a nicer experience, but you can easily fall back to a regular email, client and if you see see an email first, then I can read it still. Yeah.

D

And there's there's the whole UI kind of component to this. We're not really going to get into that's but yeah. That's the theory, obviously not easy.

D

You know you know trying to attack the the closed source or the closed models of messaging, and you know as a company-wide initiative, that this is what we think is being pushed to look at, and so hopefully people in this room would find this a worthy goal, something and again, and the idea we're using all open standards or trying to and no introduced new standards that we can use that everybody can use and it will get to here in a second, the submission token part was pulled out of COI, because it's not specific to this initiative.

D

It's gonna help out the initiative, the chat over IMAP, but we feel that the submission token idea can actually be used for any email delivery in general. So that's that's. Why we're kind of bringing this first- and maybe you know down as this coy stuff gets a little more fleshed out.

D

You know we'll see how how that is. That's a yet that one says it's a little more!

D

There's so many moving parts right now, so I think that's why you know you don't really want to get too much into that, at least in this group, and this conversation at least today and I- think the next slide would be. I can look down here yeah. So this is the how and then this is. This is going back towards us. The the submission token idea.

D

Our general idea is incremental improvements using the existing standards and until overlay on the infrastructure, with the idea that you always can fall back to what exists today, which is there is a delivery path. It may not be real time, that's real time as you want it to be, but you have a delivery path. When we, you know, we have an email that we know and love that exists today, and so the idea of submission of tokens is to somehow distribute a limited privilege token to allow direct delivery.

D

These tokens are between a sender recipient pair, so if you're sending to a group, each each sender recipient pair has a different token and there's two different token trust levels which is temporary versus permanent. The idea here being that the temporary token is something you know you're sending out. You actually don't know if the other person is even going to respond to you in the beginning and then once you initiate kind of a contact or a delivery connection, you can have a transaction. We have a more permanent, more secure token and I.

D

Think there's one more slide, yeah. So the question it are.

G

The tokens bi-directional or or if you have a conversation between two people, are there two tokens if.

D

You have a conversation between two people, aadmi having a conversation with you and one other person. There would be two separate. Oh yes,.

D

For justice people you would need well the way that the way that works is I would send you a token, and if you are aware of how this works right, you implement this. You come back to my server and say: with this token: I can deliver. You know I, you know, I have a token I have access, and then then you give me a token during that.

H

Yes, sorry, yes,.

D

Took a little while to get there but I think we got the end. So I didn't give my name Jim.

G

Fenton, okay,.

D

Yeah, probably I could have put some nice graphics up here, but I think.

D

By anyway, so bringing it this here today with the giraffe, that's out there, which I screwed up in terms of posting to the wrong place and putting the right sort of copied and pasted from the existing and existing document, but really we're looking at, is there something, but we're obviously missing here and then the people? You know people in this room that do this every day. Is this something that even makes sense I mean we think it does we've been thinking.

D

Timo can actually speak up more if he, if he we get this incorrect. This has probably been been playing with this idea now, since at least last summer. So this is not something. We've we've come up with is something that we've vetted pretty pretty heavily internally, and so the draft that we have that we've posted and is I, think I actually reposted it. So it should now show up under the extra as a.

D

Under the Act and the extra page, really, what we're looking for now is some input starting again try to get these. These questions resolved. I'll put this here. The first, the first bullet point here is: is we were looking at this where the teemo's I think teemo's main idea was that the submission servers, the submission service, would be doing all the work, so this would be transparent to end-users. You send it to your submission server.

D

Your submission server adds the token sends it off and then the other way around is what but I look at this as a more generalized use case, where that's that's a perfectly fine use case and that's actually the way that we're going to be implementing. But there can be a thing where you know you and your side. Your submission server doesn't know, isn't aware of this. So as a client, you should still be able to use. If you want to build the information into your client or build this, this feature in your client.

D

It can you can do this even without having to go through or wait and for us us a submission server to handle this. Yes, Alexi sorry I might have missed it in.

E

Which case I apologize, can you describe shortly? What's the purpose of the token? What's the advantages, what would you get with token versus the legacy client that doesn't use the token so.

D

Legacy client you're, obviously going to just the SMTP, so you have no idea how many hops it's gonna take or how long or how it's gonna queue. So that's one, that's one part of not you know. Well, you don't have a token. You go through the regular, the regular today's today's mail delivery. With a token you get the advantage of going directly to the external or it's to the person you're trying to send it to you, go directly to their server, and so you are delivering is essentially local delivery, and so it's a synchronous delivery.

D

At that point,.

A

I

J

Am I on yeah good, so a few questions does this token also come with a separate server address thing, because you're saying it's gonna bypass all kind of existing hops and go directly to the mail store. So this a token plus some kind of URL or I.

D

Don't know now so the IP, the idea right now is the, and this is sort of the tricky part which is down here is the auto-discovery part. Is how do I know? You know if you have foo at example.com. How do I know what server at example.com to go to and the idea right now is to use the existing SRV records because we're doing we're just connecting to the submission server that already exists. It's not it's nothing new, but it's a submission server that already exists under the current standards.

D

If it, but if it already exists, it's not the one. You would normally.

J

Deliver mail, no.

D

It would be yeah you cuz, we these days, I, don't go when I'm sending something something to somebody at google.com. I, don't go to the SRV record for submission I'm, sending it to the MX record. It's.

J

Like a second, a separate MX record, then essentially once you have token you're doing a separate, but it's still something from the DNS just separate it to the MX records, correct, okay,.

J

K

Speed Reznik um I'm trying to wrap my head around what exactly is going on here so in the normal course of doing email. I would be perfectly happy to get additional routing information to go directly to, and it wouldn't mean the submission server. It would be the delivery server I would like to go to the delivery server associated with the IMAP store for this particular user, because it'll go faster. So it sounds like there's two bits here.

K

One is give me through SRV the place that I should connect to directly write an SMTP land and the other is queue-jumping right. I, don't want to be cued with the rest of the mail. I want to bypass that and jump over. The queue are, which one are you doing or are you doing both or I? Would.

D

Say we're doing both the idea is this? Is it's essentially two trust model? This is a way that you know it's a trust model it's saying I have taken. You know, I decided to send you a message. Therefore, I trust you enough you're, not some random person on the internet. That can send me a message because you still allow that, but out what I'm saying is I, give you permission to come back to me and I.

D

Have you know when you come back and you send message through I might have a different trust model in terms of the way that I scan your mail spam spam reporting rate limiting those kind of things. Okay,.

K

So so let me drill down a bit. This sounds like there's few functions here and I'm, trying to separate out the functions from the the implementation. So first thing I've got is effectively a buddy list right I'm saying these are the people who I'm willing to get direct communication from correct, okay and so I'm, basically telling my server. If this person contacts me send them on three, no, you don't know need to be spam.

K

Filtering no need and we're gonna do I would say no spam filter well, but you understand eyes, given the name and I say this person's okay and they've got the right credential, which is gonna, be more than just an email address. Then they can just come on through and bleep up on. My screen, that'd be great.

K

Independent of that I want the transport mechanism to be more direct, although it sounds like that, doesn't really matter so long as I. Don't get queued right if it goes through multiple hops, so long as those hops don't take a long time so long as it gets through to me pretty immediately without being held up in spam, checking and all the other stuff that currently SMTP does that's, okay, you would agree.

D

Yeah I think I'm trying to get I'm trying to wrap my mind around that also yeah. So I think that that sounds. That sounds accurate. Yes,.

K

One of the things that I'm trying to figure out here is: why do I the sender need to keep any. The token is only my credential. Might my my crypto junk to send to you so that you know it's me? Yes right! So it's it's not a token per se to indicate I want this service, it's here's, my identification and for regular email, let alone chat.

K

I could use this to say, get through quickly and skip some of the steps that you would normally go through, because I'm me and you know, I mean that's the idea. Yes, okay and and I think. Maybe the mechanism you've got here seems a little funky for some of those features, but I get the desire for the feature and it's interesting. Okay, all.

A

We've got Neal remotely again so pop him back in next.

A

Neal you're out there.

A

Hello, Neal: are you there?

A

Yes, I'm going.

D

To put Neal off and you can come back in again see high-latency we're talking we're talking for low latency yeah. That's that's the goal here again.

G

So hmm lots of questions here, so it seems like there's one phase where you're negotiating the tokens and I can sort of see that as happening, possibly through even normal email as a as a as a bootstrapping mechanism, but once once you've done that and you've got you've got tokens I'm, not at all clear. Why? What? What does? What does the rest of the email infrastructure bring to this? That that is preferable to just having some dedicated Meccan, probably lower latency mechanism that you can just talk to the recipients, server.

D

The I think the idea here is I would love to have a better mechanism for routing and and and but but I think the problem becomes. How do you roll that out on a worldwide basis versus something like this, where we feel this is more of an incremental upgrade right for software that already exists? It's something that's slightly different, slightly more. You have to add on to existing software rather than creating.

D

You know again, like you could create a good jabber like you know, infrastructure where that it they they have created a system that does a much better job of this. But how do you roll that out? Is that something that's gonna be easily? You tell all the male providers in the world now we have to start sending messages this way or you tell mail providers. Okay. This is a small incremental upgrade that you can do to your servers that allows this sort of access, but.

G

The problem I'm having with all of that is that there's an awful lot of baggage that comes with using email as a transport for like a one-line message, I mean you're, basically saying it seems like you're you're, sending an email, a full email message for every time somebody hits the return key and that that that I mean it seems like a lot of overhead and I. Shouldn't worry about overhead, but but I am more worried about, is the the latency of that, because you know when somebody you know.

G

Typically, when somebody sends me an email message and they press Enter. There's, maybe you know five-second delay and I've got a really minimal mail server, my own mail server and there's like a five-second delay of between when when they send the message and when I get the message and- and that doesn't seem like it's gonna- be very good experience for people yeah.

D

But I think you're getting more into the application of this like check good again for the submission. Stop it alone, we're just looking for improved. You know this is this: to to AUSA's is a better trust, or at least it adds additional trust to a little exhausted.

G

I totally get the tokens, in fact they maybe talk to you offline about that. I've got some things: I'm good, but you're talking more about the real-time.

D

Of why you need millisecond sort of that kind of stuff, right, yeah and I think the answer to that is: maybe your use case you don't or in your server. You don't want that sort of that load yeah. This is as this is a question.

G

That that is coming up with, and it.

D

Really even queuing things that.

G

I think Pete was asking or and.

D

G

So forth as well, just how are you going to be able to use this as a yeah yeah again.

A

Hello Neil.

J

So two questions first I'm still not entirely clear how this is going to make be delivery significantly faster, because you're still going through kind of same SMTP providers, don't try to make their systems slow. Is the only thing that it's bypassing kind of spam checking that kind of thing, because you know you're going to have probably load-balancing SMTP service or something on the outside.

J

Anyway, you probably not gonna want people to bypass that I can't see how you're gonna make a significant impact from this on the delivery time, because it doesn't seem to be chopping off enough essential.

D

Question would be.

G

We potentially see.

D

A speed-up in the delivery time I mean it our experience that, for example, the customers that we have very large customers. You know they actually have mail queues that don't even exist like if you send through regular channels the mail queues, don't even exist at the organization they're there in the cloud somewhere AWS you know, and then it goes through the you know: the provider, the the state it the anti-spam provider. Doesn't you know they live, can put somewhere completely separate and yeah.

D

Obviously, you're still gonna have some of that potentially in the background here, but then that gives you you know it's a choice that you have to make as a as an admin is Who am I willing to do. I have enough trust in this system, where I can do much less anti-spam or I can do something like I'm willing.

D

You know, instead of doing heavy anti-spam I'm gonna do rate limiting instead, like that's gonna, be a better where I don't have to queue messages, and then you know or there's gonna be a limit to the size of messages that can be delivered. This way I mean that there can be there's several different ways that we see that you potentially yeah. This may not be a general purpose solution that everybody can agree with, but at least it for some subset of messages. This is something that so help out.

J

Okay, so it's for when you have a off-site kind of mail filtering thing for its kind of spam checking, and you want to be able to bypass that okay I can see what, if you don't set about that, could see and I guess that she answers what was gonna. Be my second question of why you need to token, rather than just if the message is dekum signed from the sender, the sender's on your buddy list, that you trust potentially.

F

J

The domain being signed, but not the user, then why can't you just use that blob of requirement token, but if you need it to gone to somewhere different in the first place, then yes, I, guess that doesn't work right. It's.

F

Pretty rare that we're Deacon signing I equals being different from D equals.

F

Alexis's, but still you can reuse existing infrastructure and that's kind of where I was gonna go. This is Barry Neil. Are you done yeah.

J

F

I'm sort of following up on some of the things Pizza and then some things that have come up since I that the general way this this sort of stuff is working is, as Neil said, this stuff's in the cloud I'm I've got. My system is front ended by some spam filtering that I have farmed out to another company and stuff is bubbling in that way, and even internally I'm not going to expose my internal structure to the outside.

F

So if I have three hops inside my company before it gets to my delivery, server you're still going to go through all that, and so again, I have the same sort of question. I really don't see how this is going to expedite anything and I, don't see how a lot of organizations are going to be willing to allow this kind of penetration user. It would seem to me that this would expose the situation where a user could allow somebody to bypass some of my security or at least defensive mechanisms.

F

What do you have some yeah? Do you have some there? For instance, one thing I'm blah I'm trying to block by this is malware coming in on email and.

D

F

Certainly not going to allow one of my users to say mail from Aleksei can bypass that mm-hmm I. Don't think we're saying! Oh well,.

D

F

How does this really benefit the system if, given that or given all of that, but I think you know it to.

D

Me that's why it's important that there is we're not saying this bat bypasses spam, but I. Also, you do as me as an admin. I would want to be given if I find give another way to trust it to do to increase trust like Dee. Kim is fine it. You know it gets me from where you're the main level but I, don't you know, I, don't necessarily trust, but this is something where this sort of idea is.

D

I have a user that actually has gone to another user and obviously there can be hijacking of a token somehow and but they would have to you know you have to be in the domain and just be sending messages as a user, and there is some hijacking issues that that's part of some of the questions that we have to work through, but I still at least our experiences are, it does take I mean because these cues get full of junk messages and spam messages they have to be going gone through, and if this truly does increase the the reduces the amount of messages and this pipeline by you know ten times or something like that, one, you know a factor of ten that to me is a factor of ten increase in the speed that that message is delivering.

D

That pipeline is worthwhile. It's a go after.

F

I agree with that: I just don't see how it's gonna do that: okay,.

D

F

What is the actual mechanism by which these mythic messages are going to bypass anything or are going to reduce queues anywhere? Well, it's.

D

Synchronous delivery right. So when you deliver the message, it's synchronous delivery. So you have to sit there and deliver the message or not deliver the message just doesn't go into another queue. So it's not okay, wait! So messages.

F

Sent with this or not store-and-forward, this is not store-and-forward. Okay,.

D

Then I completely.

F

Think it's not going to work because I'm absolutely certain that you're not going to get major email providers, exposing the their delivery servers directly to the internet. That way, I could be wrong. I.

A

Doubt it okay, yeah, there's some stuff in the Java channel here from Ned Freight spam filtering is done in the cloud in a variety of ways: a full-on sntp Hope's, only one of the ways essentially proxies or another and says Milt up. That said what I see in typical corporate setups involves a bunch of internal hops. We try and drive a number of hops down in our deployments, but customers are surprisingly resistant to making the necessary changes and yeah. Even when they save money, they still want to cheat all those things so.

K

Pete Resnick again to answer Barry a little bit and then I'll go spelunking. So if you're running a jabber server right now with a sort of jabber extensions turned on people can toss malware across that channel anytime. They want in much nastier ways um so so long as this is implemented as an extension which causes the SMTP channel to become effectively an instant message. Channel and people agree that that's what they're going to allow on their systems, then it's plausible to me now.

K

That's that said, you've got to implement it end to end, and that means that people who are going to use this mechanism have to participate on both ends. Whether that's the playable on out is a much more interesting question. Either.

A

Way to implement in intent, you could have this channel one way and regulation to pay for the other way, but.

K

All I'm saying is on in one direction: yeah in one direction it has to be both ends have to decide that they're going to do this um so now one level higher and then I'll just drop down for a second okay.

J

K

The one level higher is, there is no way on a etf's green earth that this should be in other than its own working group. If it's gonna get going. But yes.

H

Thank you, okay, there's a lot.

K

To be talked about here and I do think you should take very seriously the comments that the token stuff and doing some sort of whether you want to call it authorization or buddy list or whatever it is, is an interesting, separate issue from the real-time delivery now to spelunk. For just a moment.

K

If we're going to talk about real-time delivery during the IETF fax working group, there was a proposed, but I, don't leave ever standardized immediate delivery, SMTP mechanism way way way back in, like RFC 821 and prior days, there was actually an SMTP send command for exactly this mechanism. It is long since gone with the dodo herd, but if what you want to do is keep it within the SMTP stream, but have magic happen, we could resurrect things like that again. I think you want to separate out these questions into the architectural bits.

K

Some of them are going to be easily solvable and useful. Like the token bit, I think some of them like getting into end participation to be able to bypass a bunch of crap, is going to be deployment wise hard, maybe mechanism wise, pretty straightforward, I.

A

Have I have one really bogus proposal to throw out here, which is that, if you're going to implement this in the most hacky way possible, you would provide an IMAP login token that allowed you to append to a mailbox.

F

Okay, this is Barry.

F

In response to feed the do you think there's any way that Qualcomm would have allowed someone outside on the Internet to come in and directly connect to some delivery server as in an end-to-end way, especially considering that the authorization to do that was generated by an end user somewhere in the system.

K

What's my old line, I really like strong men, they're good neighbors, they don't make a lot of noise and they don't smoke. So no I don't think in any way shape or form. Anybody would be stupid enough to deploy regular SMTP and allow me to go to a delivery server directly except I. Don't think that's the only mechanism for this kind of thing, and- and so if you know I, think yes, this has gone off in a funky direction, because these guys assumed that they were limited to a certain set of solutions.

F

Okay, so I think we're kind of all agreed that none of us really thinks we're going to do this over SMT. Well he's a regular S&T. What what I'm wondering is? Why not set something up using HTTP, which was designed for this and Pete, doesn't like that either, but I.

F

Don't know guess.

A

What I suggested they just gonna hit and push a method, trust I'm.

I

Glad nobody has strong feelings of this subject.

F

I, like the idea, so all right, the other thing about instant messaging, which we'll say instead of chat, let's call it it. Okay, that's a.

E

Tincture years.

F

Ago, when we started trying to standardize instant messaging and eventually wound up with XMPP and simple, we had lots of debates about the exact mechanism to handle this stuff protocols like beep were developed and stuff like this and there's all kinds of fallout from that, and we wound up with two standards.

F

Aleksei says it's xkcd a bit: you have. We have too many standards. We have to go down to one now we have too many plus one.

F

So I would rather see us try to do something like this over a protocol that was designed to be an end-to-end protocol rather than a protocol that was designed to be a store-and-forward protocol from.

A

Java, just before it comes up here and it says, you're all rearranging the deckchairs- the protocol does dot naught dot matter.

F

K

Let me strongly push back on Barry and softly. Push back on net cuz ned scares me because he's big and me right, um no exactly one of the things that I think you miss. If you go down the path of oh, we should do another protocol is that one of the design features of this proposal is that it falls back to SMTP messages.

K

So if you can jam something into the SMTP pipeline, that says delivered this more like an instant message and that extension isn't there it's delivered like an email message, that's a design feature, and so, even if the and and Berry says he'll buy that, so even if the mechanism here is I, go to my SMTP might SMT my submission server and hand it a token and my SMTP submission server goes screw this I'm, not using SMTP I'm gonna, do HTTP directly to the other end. That's a fine outcome.

K

So long as the back off mechanism is I've just submitted an SMTP message. You know if it goes.

B

Chris Newman there is a standard out there on the standards track that that's.

D

An smtp extension called deliver by which, which says I've got this message, deliver it by this time and if you can't bounce it and you can give it in a delay or whatnot. So a former coworker wrote that document that former coworker now works for tinder.

D

G

A type of note sharing it's a type of messaging.

G

Anyway, what if you just said, you know if you got a if you got a deliver by you, know one way to implement.

D

That is to yo trigger a direct, yet yo trigger it trigger more direct channel.

C

F

Is barry the the meaning is there's different semantics there, though chris there's one is saying: if you can't deliver it in a timely manner, don't deliver it, and the other is saying expedite the delivery of this, because it's some special thing so there's a very different semantics. I can.

E

Clarify there are two modes: there is a mode where you say well still try to don't bounce, it continue delivering it, but it will just track and say: oh it's it took longer than expected. There then we'll receive it. So it.

F

E

Close enough yeah, there is yet another thing. There is I've done, SMTP priority extension, where you can basically say you can tag a message saying this is special thing again all different pieces of bling.

A

As bit of stuff happening in jabber channel as well, Anil says I agree with pate. This perhaps value and just a token say the SMTP server give this priority, which the server can decide. What to do is I've, followed up on that saying that the issue is that the source server doesn't know if the token is valid or not. So if it just prioritizes anything, that's got one of these tokens on it then.

D

Everyone's going.

A

To add a token to it, so they get prioritized ahead of everything else, and the flow control at the source would be broken. It's you need to know whether the destination will accept this and prioritize it. Otherwise, it's just the high-speed bit that everyone will turn on.

A

Ned says drilling down past all the details. The essential characteristics of this proposal appears to be that the message ends up in the email message store and that the client knows that the message got there correct. Yes,.

D

I would say: correct the the yes I mean the nice at least the way they was drafted. um A nice artifact of that is that you do have delivery, like you know, in a message in a more message or Ruby, calling a chat, instant messaging, where you send a message and it says, delivered or something like that right. You know you you get that and when we're looking at this overlaying it as to what we want to use it for that's one of the things that it does get us.

E

Just sort of side note type comment: I think a lot of us need to digest it.

E

There are lots of concerns and there are lots of you know. There are bits that are interesting to some people and draw other bits which looks scary or not interesting, so I think we probably should take discussion.

E

Continue it Sam.

D

E

D

You said it's extra the right place for this no discussion, yeah and they're, not.

E

Gone wanted to talk about reach. Our Turing and I were to slightly hijack the discussion because of SMTP aspect of it. Yeah I want yeah I want to social, remove.

A

To that just yeah, excuse me yeah. It is time we thank.

E

F

Yeah, so this is barry I'll saying that, in summary of that, yes, I think it's interesting, I think discussing the. Why and and the weather is a good thing for us to do and we can beat out the how later and I agree with what pizza that it absolutely has to be its own working group. So.

A

Alright, thank you. Do we want to still look around a quarter, or do we want to jump straight to reach out a discussion now reach out all right? Let's talk about reach out of, then, we have mostly finished what are charter originally said certain at the charter up here and which I meant to do.

A

Sister, current charter.

A

Will work on updates extensions and revisions to IMAP sieve menu sieve, we didn't say pop3 in there we'll start processing the backlog, which we have basically done this IMAP for over to do and there's nothing else on our horizons. Right now,.

A

We've done some sieve. We have said we're not going to do any de marchi crap Gemma, pop3, SMTP, MTP, email format, slash mime, but we've now done.

A

Others, the yeah the quota base document as well, the EA I would see if we have not adopted.

A

But it is time to look at this charter and see where we want to go from here.

A

So I guess Alexian very probably have the the biggest opinions on what they want from this group. Well,.

E

So when I was chattering, this working group I actually wanted to include SMTP as a part of it and I got pushback from some people, because they're also you know there is SMTP- has revision of the base bag, which is overdue. There are some possible extensions, so we thought that was quite enough of work to do so and recently there was a discussion on ITF SMTP mailing list again asking for SMTP related working group, so I'm thinking we publish, which are SMTP related working group.

F

This is Barry. Why do you think it needs to be an SMTP working group rather than a continuation of email? Stop working group like I mean that's something I think we could do in this working group without I mean we'd have to reach Artur, but why would we need to have a separate working group and so first of all, okay I would not want to do smtp without doing message format as well, so we're talking about up updating, 5321 and 5322 right, and then we probably want to do mine.

F

So we have this set. I would want to do the first two first then attack mine. Do we have a lot of work? That's coming into this working group that we would want to separate from updating, 53, 21 and 22 I.

E

Would like to be able to work on SMTP extensions without requiring a wait. A couple of years.

E

Well, considering some people involved in this, this is a very practical reason and I'm I'm I'm. Sorry I cannot be more back than this.

I

F

The big deal with adding another work: now there isn't so you know I, guess: I wouldn't object to that I'm. Just it's a different I, I guess. One of the things that is is that it's a it's a question of what we're looking to do with 53, 21 and 22, and my view of what we want to do is move them to internet standard and minimize the changes we make. We have a list of things that we know we need to fix in them.

F

I don't want to start I, don't want to drum up all of the people who want to put in extensions into the base SMTP or you know more other changes beyond what we know are glitches and that's what's going to take years, if we want to do them and I'd rather not.

E

Some of the reasons why I want to work on extension- and this is slightly hypothetical in the sense that there are a couple of extensions- were not necessarily sure whether they're good extensions I would like to be able to exercise SMTP, extensibility and being able to actually publish extensions in this organization where you don't have to have ten years of experience. To do this.

E

The last two people who were able to publish SMTP extensions, I think like Marie and me, and we had over 10 years of experience within this organization, I want to make it easier for new people to do this. If they have sensible proposals net.

A

Has something he wants to say to mark which is: does the update of RFC 5321 need to be coupled play up that of RFC 5322? Yes, there are a few interdependencies, but can't they be managed I suppose.

F

They can. This is very neat, says. 53:22 is a lot quicker and I'm, not sure about that, because I'm thinking, if we are, if we go with the model I'm talking about 453 21 I, think that should we should be able to get that through fairly quick, also I, don't know, but I'm a hopeful sort says Pete. Yes, I am anyway I.

F

I was not suggesting that we shouldn't I I'm, not suggesting that we shouldn't work on extensions while we're doing that so I guess that would say we should work on extensions here and have another working group, progressing, 53, 21 and 22 I mean I, guess, there's no reason we couldn't do 21 and 22 separately, I, I, guess I've said what I want to say.

A

All right, what's the action that comes out of this I.

E

Think we're talking about plurals so.

F

Florals, only in the sense that we need we reach are during this.

E

F

ah So that's one charter and then the other charter is for the working group that would progress, SMTP, message, format and mime in some stages over time.

E

So I think Maria and I will take an action item off atha draft some text or you know organize other people to to drop some text.

A

We still have nine minutes left I.

E

Might be able to take talk about what, if you want yeah.

A

E

So I looked at all last time we were met. We there was a interest in quote extension and I. Looked at all the lists of my expired draft and found a 12 years old document which was about quota, so it was also in XML format. So I was very pleased about that. So document was actually co-edited with Dave Cridland and he did some quite a lot of good checks. There I don't think he he wants to be an actively involved, but this is basically a minor revision of this document.

E

Next slide right. This is the story yeah and before we wrote the document twelve years ago, there was another attempt earlier on, which was more complicated, where we added lots of extra commands.

E

To remind people what RFC 2020 87 does is, there are three commands. There is command to discover quota route, which is a collection of mailboxes. Quote applies to for a mailbox. There is a way of querying quater route and there is a well way of setting code road information, early attempt with Dave :. We added lots of commands like delete quarterly squat rules and make it more flexible, and then this was great, but slightly over engineered.

E

We decided to scale it down back, and the latest attempt is the minimal set of useful changes and extensions to the existing RFC I. Believe.

E

Right so I'll see 2087. It was a bit vague about what what applied to it gave a couple of examples. It never promised that there are any anything more than examples and some servers implemented them exactly as sort of implied by the document, but they will never really document it.

E

So this draft standardized, 2/4 type resource types that were in the original RFC properly documenting, is basically storage, which is a total sum of all messages in all mailboxes covered by the court route or a number of messages, and it adds another one which is a list of mailboxes allowed per user quota route. Again there is a extension to advertise this so that clients can actually figure out connect to the server and see ok, I understand this quota resource type. I can do something with it and you know, depending how generic they are.

E

They might be able to present some UI for the other ones. It doesn't understand or just ignore them entirely.

E

Then, additionally, there is new response code which says over quota. You know, if you do this operation, you are getting very near the quota for the current mailbox or you went over quota and then a couple of status command extensions which basically, if your resource, what resource type is messages number of messages deleted messages will say if you expand this mailbox.

E

This is how much you can recover to get. You know toward your quota and similar for this deleted storage. If that.

A

Colonel box thing, rather than per quarter it yes,.

E

Status is per Mobile's.

A

Right, there's no way to tell from a quarter route the total amount and you.

E

Have to do it for each mailbox yeah again, this is one of the this is actually one of the open questions about. Are the only useful if you implement certain quota resource types and like next.

E

Yeah a few questions in the document: I, don't know whether people want to discuss them now.

E

One suggestion was that you can say get what a route from a non existing mailbox mailbox to be created, and then it will say well, if you create this mailbox, it will fall under this. What route.

E

Yeah, with over quota, there are cases you know you might have. I'm I have one network selected for the at the moment. It is defined for the country selected mailbox, but if you're copying and moving message to another mailbox, do you want to know which mailbox it or what applies to again this you know. If people are interested solving this case, then we can extend this index to maybe convey this information.

A

And I had one additional thing that I wanted to suggest there, which was doing a quota for annotation storage. If you support the per message annotation, then at the moment that doesn't have support for any quota and we have a an extension within Syrus for doing a quarter on that as well.

E

So I don't think she want to own this document. I did it because I already wrote it once and I thought it was close enough to what people want.

E

Is anybody interested to correct it.

E

Sure, absolutely you know that.

D

Discussion entered the list, Michael is ours. The mailbox thing is really nice fact that serendipitous came up today. Early this week, like we had somebody asking how many mailboxes do. Our users have it'd be nice. We would have to track it for this right question that did come up. We were reviewing this. Does anybody use set quota yeah.

A

D

Okay, so we we were yeah, because it's kind of a weird that, like it's set core, is only useful if you're an admin and it's you know for this I suppose.

E

Yeah but well that's a good point. You know we can possibly make it separate capability. You.

F

E

People really don't want to expose this side of thing, but you know yeah.

D

And then there the final final point was: there was discussion about physical storage sizes versus RFC 822 sizes. We actually would love. We just did RFC 822 sizes, because that's there's there certain storages. We have no concept of what physical sizes are, because we don't know how the.

E

Intent with storage was I think it's. We had similar issue with another document recently it has to be no less than the total sum of messages if you store, if, if you round up or store extra information, that means Counting, you can edit them, but for the most part, is basically yeah at least the sum of messages. So if you want to just return the size of some of messages, that's fine basically.

D

Chris Newman, so the two issues that maybe may merit some thought with relation to this spec are: one of them is quota by type where different types of messages in the mailbox fall under different quotas. For.

E

This resource type quota resource type can be by message type. You just define a new resource type right.

D

So yeah that III need to go see how what the spec says compares to what's implemented in our server. Since that's implemented,.

B

In our server well again,.

E

You know if you have a specific message: types like I, don't know. Voicemail.

H

Voicemail might be.

E

We'll just define another one of these and be done with it. If it's something more complicated than you know, yeah.

G

I mean I personally lean towards you know, let's just find one for voicemails it'd be done with it.

E

That's fine, in fact, one of the action items I think on the next page is create a on the registry for resource types. Then it will be easy job. Okay,.

D

Now the other issue that merits and thought is places where they have one quota, which covers not just email but other resources.

D

For yo, so a given user has one quota, but you know if they say upload a word document into their kids. There they're.

B

D

Counts against.

B

Their one quote.

E

That's actually related to required conserver if the server changes that if the quota usage is counted toward other things, somehow server can send you an tagged responses. Well, that's open question. I'm, sorry have load word document you might have I. Remember.

B

Probably don't want to do that because, because the yo.

D

When you have this situation, where you have a master quota and the imap is one one thing that contributes to your master or quota and usage, you you, your IMAP server is not the source of truth for that information. I can.

E

See both yeah this one. Okay, can you send this to comments to the main in this? So I can keep track. Okay, but I believe you know both of this can be satisfied. It's actually pretty close to what you want. Okay,.

A

Sounds good all right, we are now cutting over into the J map meeting, which starts right now in exactly the same room. I believe that we need to adjourn the current meeting and and well do you want.

E

To know how we accepted this item provisionally or.

A

At last, okay, and and said that we accept a quote of this document to be delivered. Okay and.

E

Chairs I assume happy with me with my call to coordinate.

B

This is the blue sheet for extra. Yes, it is not blue. She for J Matt.

A

B