Internet Engineering Task Force 101, 22 Mar 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF101-EXTRA-20180322-0930

Description

EXTRA meeting session at IETF101
2018/03/22 0930

https://datatracker.ietf.org/meeting/101/proceedings/

A

Do you want a Java scrap since you're sitting right here, yeah, cool and I'm, happy to take notes? I, don't think we'll have that many decisions to note so good morning, everybody- probably particularly good morning to Ken who's up at 5:30 a.m. over in the US, welcome to the extra extra work. Now it's not the extra we've only got one extra working group. This is it on the Thursday morning. We have here the note well updated. I did actually check.

A

There are two copies of BCP 25 in there for different reasons and they're, both actually I mentioned in basically 25, so I think that's actually real. You obviously can't read that slide, but you can easily search this seam on the Internet. All right are you doing well, I've made it very, very small. This is the rough agenda. I've put together absolutely happy to rearrange move things around whatever people need. Does anyone have anything they want to add to the agenda? At this point, I've already noted that Barry's proposed new draft is in there.

A

Do you think five minutes will do for that excellent? Shall we begin I?

A

Have there are only six open issues in github I closed the two regarding unique ID, because I have now written a draft for unique IDs, I promised and there's been a bit of backwards and forwards on that.

A

So we have issues on saved a special use, something I created to put an informational document about how thesaurus replication protocol works, because it's designed to replicate the IMAP data model and it might be a useful thing to have- and there are a couple of issues about 64 bit or 63 bit extension I, believe all of these should be covered in apart from the size, replication thing there's space for each of them under their draft. So we probably should speak about them. Then sorry.

B

I guess we do. We have a separate item for the 64 bit. Yes,.

A

B

I'll wait till then.

A

All right, so these are the existing drafts that we have sitting in our queue in various state. There are six of them so far and we have three more proposed coming down the way. I don't have enough experience to know if that is an insanely large number to have in a working group at once or not, but some of these have don't have much work left on them.

C

Yeah I'll have to wear multiple hats here so with my ID hat on I think it would be nice to get one or two of you queue yeah, because I suspect that list my right instead stateís eyes are done, maybe see especially use as well. Yep.

B

This is Barry and I. Think the one I'm proposing is basically done on contact.

B

Isis I think we should leave the imap Rev for after we get a few of these off the queue before we start on them and then concentrate on that anyway.

C

Definitely have a conflict of interest because I think you can probably speaking as a participant. You probably can work in parallel, but you don't need to be done with it all.

A

Right wait all right! So I now have slurred speech this. So since Ken's on line, are you awake enough to start talking about your ones, Ken.

A

Here we go all right, so there's the first slide and here's Ken.

D

Waiting for the slide to come up, can you guys hear me? Okay, yes,.

D

Hey beautiful both of my eye, my eyes aren't turns, are concave ago. Okay, so I'm atlas right. This is very, very similar to an existing RFC, which does the status extended the list. So all this is is a new return option for extended list which will return a my rates response for each of the existing list of all mailboxes that come back in the list response itself. The benefit of this. It reduces a round trip for each of the mailboxes that a client would want to get the rights law next slide.

D

Please- and there is a very simple example of what it would look like, as the only interesting bit here is, is the last mailbox bar which is not existent. You obviously do not get a my rates response for non-existent mailbox.

D

That's pretty much.

B

It shell questions, and this is Barry. It strikes me that retrieving your rights could be an expensive operation relatively speaking, is there any discussion of that in the draft? I, don't recall, I.

D

Don't recall off the top of my head either Barry I would say: I don't have to take a look so I.

E

D

There is a mention of it: I do not recall to what level.

B

Okay, because I just worry that if someone had access to ten thousand mailboxes- and you did this command right- it could be very expensive and it might be worth calling that out so I.

C

Was thinking about how I'm going to implement it in our server and I suspect, I will probably do least return my rights and specially use at the same time, because that's a note that you know- and this is list is very small- so that's probably the mailboxes I care about and everything else I will do as as the user selects them, but no I'm not suggesting that this document needs to mention special use, but I think it's probably worth discussing.

C

Don't other comment. I have, is you have the M right, showing an example which I think is annotation extension or something yeah? Yes, so I think because it's not part of standard time app, ICL, it's an extension to it. I think you should point out and in a comment saying that well in case people will try to actually analyze your example where this is coming from obvious.

D

Remove it'll, actually it's probably simpler that way: yeah.

A

Yeah, that makes sense. Obviously this this has been implemented in the cyrus server, because this is where it came from it's used.

A

Certainly, the reason I asked for in the first place for fast mail is that it's used to display whether a mail box is writable or not through to the UI and that's very useful to get for all the mailboxes at once. So we we run my writes is very cheap in our server and we just run a select or mailboxes with my right. So we can keep the display up in sync, yeah.

C

I think in our implementation, we in a web mail server part of our implementation. We open drafts and sand folder as soon as the user logs in internally. So you want the writes on these. You know that's sort of again so to know whether well do you have this folder and you know what kind of writes.

A

All right so I guess next steps here, African to take those two suggestions, probably produce one more draft and then probably take it to working group last call after that. I would think I can't say anything else. That's likely to be an issue here, but.

D

Back to Barry's comment: there is a one or two sentence, paragraph that does say that if the server is unable to look up the set of Rights that the server may drop, the my rights reply, I, don't know if that's sufficient or if you to expand on that. No I was looking at advice to clients to be.

B

Careful about how they use it: okay, fair.

F

Enough did you want to well, we have have a room poll for people interested in implementing it just to verify the spec house, sure that seems reasonable. Does Chris Newman by the way.

A

I'm taking notes sure he seemed interested in implementing this so.

B

Few people in the room, I don't know that only the room is going to be too valuable, but.

E

Who's gonna put it in a client. That's the.

E

B

Usually less worried about servers, because you know we got a bunch of them and they all seem to want to implement everything. So.

A

All right anything else on this, so should we move on to the next one cool all right? Thank you, Steve extra FS, they say well, we've got a TA Ken.

D

Okay, this this draft essentially, is a way to copy a store, a copy of any outgoing messages that get generated by sip actions. Obviously, the name SEC came from a non-standard header that some some email clients use right now that the draft stipulates that it's only compatible with vacation and notify it is specifically not compatible with reject any reject. That was basically because it violates the SMT principle of either message is delivered or bounced. This would break that, obviously, by storing a copy next I. Please.

D

Is the next slide coming up? Yeah go yeah, so.

A

That there's a.

D

Very simple example: basically, the tagged argument takes a mailbox name, so in this case it's added to the vacation extension next slide. I did make this work with the existing file, two options that make sense, so those additional options would get placed in between the FCC tag and the mailbox name. So in this case you could use special use, create and add Flags I, don't know if that is too complex for this draft, but it is a flooded in Cyrus, so it's certainly doable that.

A

Was Barry : looks good if it didn't come through on the microphone.

C

Okay, thank you. Alexey III, do have possible concerns in a sense of I'm I need to double check. Whether save back allows basically you're sort of trying to say that you have the grammar structure, that you have options to existing tagged parameters and then you are using other tag parameters names with the same syntax here and a question number one can can they be repeated with and without FCC in the same statement? I, don't know the answer to this. Maybe it's a stupid question and if they can, whether this is actually allowed.

A

C

To the serve spec, so it's not a showstopper, but I have a concern and I think I have a work item to double check whether this is okay or not, and I would like to hear other people what they think about this. This.

B

Is very I'm not sure what you're getting at in this case I mean these are parameters to the FCC action, so I don't see that you're not using special use as an action you're using it as a parameter. It's.

B

Well, there's no nesting of parameters done here. It's there is there's the special use parameters to create parameters, the flags prep.

A

Yeah FCC is a parameter to vacation and then it has.

F

Something yeah there's definitely a level of nesting, yeah.

B

Yeah I think it's okay, Alexia's gonna check, so I'll shut up.

A

Apart from that, are there any other issues we should be so again, I think this is probably resolve this issue and then take it to working group last call another.

C

Question I'm sorry, I haven't read the mailing list. There was a discussion about reject whether it's actually, okay and I think mad commented on this I. Don't know what the other line.

A

Was happy with the resolution that we say it can't be used for reject and nobody else wants it enough to argue against that. If there's a demand for it again in the future, then we could do yet another extension to allow it for reject, but.

C

You know intermediate message in the thread saying that actually, maybe we should be allowed, because this is internal to the server anyway, so I didn't read the conclusion. This yeah.

A

Nobody really needs it for that. The thing that that we definitely wanted for a fast mail is if we send out a vacation response. We want to have a copy of that. If we send out a notification response, we want to have a copy of it, so that we have that history of everything that's been done on the user's behalf, reject we do at the boundary anyway, so and I think a lot of sites will be similar. This.

B

Is very I agree that we should not allow it and that anybody who wants to have it will implement it anyway and ignore that so I don't care I'm, actually.

C

I remembered another issue: the way how notify framework was written was very carefully. We were separating method for notifications from the general framework, so notification methods is keyed of the UI that you are using. So you can use notifications to you to generate XMPP some messaging notifications or you can use mail to to generate emails and the way the document says. Is it doesn't talk about this? It just assumes that you can always archive now it might be a wording issue or clarifying that it only applies to when you use mail to notifications.

C

If you want to talk about sympathy notifications, then you have to say much more and say: what exactly does it mean.

D

Yeah, so Lexi you're, uncomfortable with the hand waving I, did saying that it's up to the server to determine how to save a copy or, depending on the notified action. Yes,.

C

You effectively you, there was a deliberate decision discussing in notify framework that there are different sort of plug-ins for different types of notifications of different protocols, and they will describe requirements for how messages are generated, how they look like and stuff and you're deliberately your blessing over it or whether you are omitting details or you are violating this separation.

C

So I think you probably need more text either way. Okay,.

D

If you've got something in mind, please open an issue for available either option.

C

One you just say this only applies to email notification framework and you had a reference to secondary which describe the email notifications. If you to use it for other type of notifications, you say you can use it for the tab notifications, but then you need to add details about how what does it mean?

C

A

I could see text that said something like if the server supports file into F say say for this type of notification, then it accepts a civil script. Otherwise it rejects. It is incompatible with FCC and if it, if the server does have a way to take a copy of it, then presumably that's what the user was asking for.

A

C

Using XMPP as an example, if you want to allow for this, even if you don't specify the exact message generated because it's going to be effective, XMPP in message converted to email at least say so, you know this is still better than just. You know there is nothing there. Okay, so.

A

It needs some more words for that, so maybe this one. We should finish that and then come back for one more check that everyone's happy before we cool.

A

All right moving on the next one is 64-bit alexey welcome back thanks, Ken.

C

So this started as a very simple draft that, oh, why don't we make? You know message? Size is 64 well 63 bits and just update the grammar everywhere and then Chris and his email with lots of operational issues which I think about it is I. Intellibid I just haven't thought about this. So.

C

I'll invite Chris to talk more about his issues.

F

Okay, this is Chris Nilan, yeah I, so one of the issues was this is an implicit format, change the mailbox for implementations that assume 32-bit sizes and that's a you know. That's a lot of code changes. Often you know in in Syrus based servers. That's a new mailbox.

F

You know index format which has which can have migration issues. You know I mean it's supposed to auto migrate, but you know there you go, does mean it's perfect, so I'm curious, what's driving the beo before yo I'm, probably not interested in implementing it, but I'm curious. What's driving this is there you know? Is there actual interest in implementing this for some particular class of application or something yeah? They the.

A

Question I saw in your stuff and I think is even more important than this, because the server can implement it or not, is if the server supports it and a client that uses 32-bit storage for fields that are now 64 bits. The client can't really opt out because the server has data in there, which it can't represent over 32 bit anymore, so the client, the client, will fetch something that has a number that's too large for the space Astoria. There's.

B

An enable so what happens to what.

A

Happens to a message: that's too big right! Well! Well, that's the question so.

C

The coin has to obtain into syntactic extension, so the server will never send 64 bit sizes to all clients. The question happen is if a message is more than 4 gigabytes. What should the old client get in this case, and it ain't.

C

Yeah tombstone or maybe responding now to the fetch command.

B

No we'll break a lot of clients, it'll just break they'll, just up everything so well.

C

It can be I'm tagging now, but you know even.

B

In tag no I think, is it really good? Oh I think you need to either eliminate the message which I don't like or send a tombstone. That indicates it's an empty message or something at least the user will.

C

G

B

Its head, but there'll, be an indication that there's something funny there.

B

Is there really a use case for this, or are we just saying why not.

A

That's a question for Lexie who needs it? Who wants it? I? Don't.

C

Actually have a good answer to this.

C

Well, I was I, didn't think of this I thought it was a good source to the community to document this, but I personally, don't have a use case. So.

B

Then this is Barry I recommend that we park it and you know it can. It can just be an expired draft until somebody says. Oh look, we do need this because it's just too complicated to put out for nobody who wants it so.

C

I I see you know, the question is, we should probably ask, does good guys and saris fast mail camp? Well, you can talk for yourself, I, don't know if you have use case, but we should probably a staff court because they commented. I again. I just you know I can ask privately or publicly saying you know. Is it.

A

Just you think it's online from dovecot, yeah and well he's packed manning across if he wants to pack man across. We have no urgent need for a fast mouth. We we would possibly implemented, but we wouldn't be in a hurry to and we wouldn't need it Stefan did. You want to comment your next anyway, so it's a good time to hop in the line now.

A

A

Dovecot supports sixty four-bit internally, Sean Leonard, says I, don't think 64-bit should be parked. Reason is because if the implementation tail is long, at least by having the option out there.

B

All week so he's finished. Well maybe it was a dot dog. It.

A

Gives developers time to prepare okay.

B

So this is very, I I, don't think developers are going to be because this is sufficiently complex to put in I. Don't think developers are gonna, stick it in just in case just because we've published it so I. Don't really think that's going to give us a benefit.

H

Jenkins can use clarify in dovecot what they support. A 64-bit is that 64-bit message, sizes 64 bit you IDs or everything, or what.

H

H

A

Their comments to come through on jabber.

B

As long as we're talking about that, I would be more willing to support 64-bit you IDs. Unfortunately, that's that requires the same mode. Switch that that this does so yeah.

A

The Stephan says: I, don't know, Jeff might.

A

C

Don't you dance in the room, I think that's fine yeah.

C

Yeah person am okay with parking it if there is no immediate use. That's fine.

A

All right next up is status, sighs.

A

Yeah jeff says in dovecot: your dates are 32-bit, but message sizes are 64-bit, yeah I.

B

Guess Stefan appears not to have audio or speaking times.

A

All right, you're.

G

Honest yes, so what should I say? Does anyone have any questions about this draft.

B

I think the second bullet is appropriate.

A

So status size is I. Think everyone agrees is ready to go to working group last call. The one clarifying question that came up was what the size should be and I believe everyone's agreed. It should be the sum of the RFC r2 to dot size of all the messages that exists in the folder, which matches the example in the draft I. Think.

G

The drafts also says this explicitly so yeah.

C

I read the draft I think it's ready cool.

F

Chris Newman are just just vetoing: I I wanted to ask: how would this interact with the email address, internationalization extension? So what's why I ask that is once you implement email address, internationalization you've got two different classes of messages in your mail store. You've got the e aí ones and the traditional ones, and if you downgrade the ái ones, they have a different size.

A

So what did you serve a return when you asked at fire CI to two dot size so.

F

A

F

Way it kind of has to work is if the server does downgrading it has to depend on whether or not enable enable the the utf-8 command has been issued by the client, so the server will actually have to store two sets of sizes for the new messages, so I'm wondering if it would be helpful to just have a little implementation note about that. So when people are doing this, they can think oh wait. You know, maybe I should have two slots rather than one in my data format, I.

C

Agree on this I agree on the note. The other thing that occurred to me sort of related to this is we do have metadata that we store with mailboxes. If this value is used to calculate quota or inform users about quota, we can possibly say that the server may return value which is bigger than the some message sizes. That includes the other information in this case, if you support utf-8, then you might just count both of them. As you know, it's metadata about happy to.

B

C

B

This is very that's what I was going to go somewhere related to that at the white. Why? Why do we need this, and so do we do? We need it to be exact and if, if it can be, if it can be a little larger, maybe so it's an estimate, it cannot be smaller, but if we can say that it must be at least as big right now, I was just looking at the text right now it says rapid scrolled funny.

B

This must be equal to the sum of the values of the RFC 822 dot size, fetch Maps data item of all the messages in the mailbox and I suggest. We say this must be equal to or greater than and then explain why it might be different and then it would be happier.

G

Stefan I must say I'm not too well-versed in the in the internationalization part of I'm. Up at the moment, you still haven't implemented that but I see what you're getting at I think. The original reason this exists is webmail clients, so those have more trouble if they want to establish how big the mailbox is. So if that sum is a little larger than the actual value I think it's no problem.

F

Okay, one other issue which I don't think yeah I, don't think we need to worry about. I just want to make sure people are aware that this issue is out there is that you know sometimes people yo-yo when there's a quota on usage, that quota isn't restricted to just mail. It covers other sorts of objects which may not be stored on the mail store and so there's actually an integration problem with our current specs, where there isn't a clear way to have sort of an uber quota. That's that's!

F

That's integratable, with with other services, I'd be tempted to set that.

A

Telescope III for the spec. What we're interested in is a way to avoid having to count the sizes of all the messages to know which folders are big right. I did yeah just I I agree.

F

With that, but just wanted to point out the issue yeah. Thank you. This.

C

Reminded me, though, the quarter spec is really old. Spec I think it doesn't even use the latest: hey B&F, syntax, it's old, yes, so so as a question, maybe this working group wants to revise this. It's rate the rise quota and I'm sort of half volunteering Dave, Cridland and I had a proposal like six eight years ago, which expired, which was trying to talk more about different Quatro routes and quote usages and other things. So there is something to build on. If you want to do this interesting.

A

I I'd say probably, let's clear a little bit of the decks first before you add another thing in, but yes, I must.

G

Be part of rf2 right sorry about that, it could also be part of it I'm at four F right.

C

Let's talk about this later.

A

We have a couple of additional quota types in the Tsarist server. We have a quota number of folders so because that's one resource that doesn't count towards your quota see can't create an infinite number of folders. You can set a maximum and you can fetch X number olders extension. We have a quota for annotation usage separately to mail usage so that the number of per message, annotations and per mailbox annotations gets counted as well again, so that you can't store infinite data and something that's not being counted.

A

You need limits for everything that a user can do or there's an abuse vector.

C

So for this document, I suggest there is, you know one or two sentences as per comment from Chris and the discussion that followed, but I'm I'm, almost tempted for you to start working group last call now and we just take it as a one working group. Lascaux comment: okay, do this so that we can get this one quickly, I think it's really! You know in a good state, excellent.

A

Everybody is happy creases.

I

Happy with this we've got a working.

A

Group last call with with a comment saying greater than.

C

So hint-hint chairs can send the working group last call message today. Yes,.

A

Hopefully, today it's a very busy day, all right next up for we've got here, Stefan save date. Yes,.

G

Well, this one open issue: there lets me discussed on the main this and it's what to do. If we have a mailbox a text duck, she doesn't actually support the save data metadata yeah to actually store it so and if you have search queries that involve that particular save date, what to do, and also if you have the multi mailbox search extension, you can spend multiple mailboxes with your search and if one of those of a few of those don't support, save date, what do we do with the query.

C

Right Alexi here, I, don't think client will cope very well. If you fail such exam.

G

C

An extension- and there is no- there is no signal coming on mailbox open that you don't support this, so the client cannot actually know without trying so I suggest either make the fairly silent or the way for the select to come back with the you know, I, don't know no safe date or something factories that the response code or something like that.

G

C

Here with naughty mailbox search, you know, if you just do it silent or there is a way of returning response code saying well, some mailboxes didn't support this, but I carried on anyway. So I'm.

A

Not speaking for myself, I'm, not too worried about the multi mailbox search case, I think that's unlikely to be the kind of place where you would need this.

F

Yes, Chris Neil and I we actually implement multi-valve our search and, and so my preference would be. If Dell we don't have any mailboxes that don't support safety. We've got a proprietary, a private extension for save date. So it's good, it's being standardized, so I'm all all good with this, but if the way I would want to handle, it is if a mailbox didn't support, save date, I treat that query term as equivalent to internal date. Yes, so that is yeah.

F

That is that's and whether or not you get signaling of whether about a mailbox supporting save date. I, don't really care I, don't I, don't see anything useful. A client can do with that signaling. So I, don't care. If you signal or not, okay I can.

A

I'd like to tell you the use case where I'm worried about this, which is that you do an auto purge after one month on a mailbox. Si save dat query and if you do that against a mailbox that doesn't support, save date and messages that were moved in there, but we're old will disappear immediately, which is not the intended behavior I.

G

Actually heard from the main programmer of Java that in some cases we actually use the current dates at the moment where it becomes relevant and then it's safe to cash. So there's.

J

G

Store and yeah, that's you, how applicable that is to other systems, so.

B

The client figures that out when it first sees the message ash, was that that.

G

Moments the defect, the second storage doesn't support it, but the cache will and then it will be in the index.

B

So this is very my view of the end-user using something like this would be exactly with a multi mailbox search. Where show me the messages that got stuffed somewhere yesterday, so multi mailbox would be an issue, but I agree with Chris on how to deal with that, which means that you'll you'll break that case you know, but you won't find all the messages, but that's the way it is yeah I.

C

I think the client needs to know whether it was broken for any of mailboxes I. Think that's what we're arguing, because the client might potentially do something else or decide. You know if this is Verge messages from the last months. Maybe I'm not going to do this.

G

The Younger is fine.

A

Because sorry Stefan did you want to finish well.

G

We had an initial proposal to return an additional response in The Selecter reply right, so yeah.

J

I would indicate.

G

It's not supported, or it is supported, not sure which one is better, but.

A

I'm really happy with the amended Libre Office is crashing on me right. Yeah.

F

Why, I suppose, just for completeness, you could have a search term saved eight supported.

F

Which would allow you to construct searches for any case that that is better than the two things I came up with yeah. The part I mean I'm, suggesting this I'm not convinced it's a good idea, because.

F

If the problem I have with it is, is clients tend to not do this sort of thing right in yo, III I, think I mean I, don't consider it really a hack? It's it's a! It allows you to construct searches that that do what you want, based on the presence of the extension. Ok.

C

Can you elaborate.

C

F

Sorts of boolean flag, so you know, if you want to do if you want to construct a search where it's you know, uses save date or internal date, then you you, you mix this flag in with boolean logic and have the search covering both or if you want to do a search, that's for purging, then you mix this flag in and with current date.

F

Okay, I'll send that to the mailing list, so we can consider it. Okay, yeah.

B

I mean I, maybe I kind of like Chris's idea. This is very because it solves the problem of the purge thing. You can eliminate the mailboxes that don't support it from your purge. It doesn't solve the problem of the use case that I had where I want to see what I added to my mail store in the last two days and I'm, just not gonna I'm, still not going to be able to know about that from them. If there are some mailboxes that don't support this feature, but that's okay, all.

A

Right so I think. The conclusion here is: let's take us to the mailing list. This there's a favorite of discussion still to have yeah.

C

I'm, just thinking about the use key so from again I'm making some some assumptions here, but I suspect that you're probably going to support it for your personal mailboxes or none of them or, for example, all shared mailboxes or none of them I think so you're unlikely to not support it for trash, where it's probably the most useful right. So, but probably we should have a text about this saying that if you choose to implement this, consider this carefully.

B

I yeah I kind of wonder we're speculating on this dude. Is there really? We really expect servers to be to support it for some mailboxes and not others in the general case and okay right, different namespaces, but that's different from different mailboxes in the same namespace and I. My view is that I don't think any server is really gonna support it for some mailboxes in what in a given namespace and not others, so I think really what we're thrashing about on something. That's not going to be a problem in practice. Yeah.

A

So the main issue is how it behaves in the case where it's not supported, and we do need to have both a well-defined answer to that and and a way to detect it yeah.

A

Cool, so this will go to the mailing list and for further discussion.

A

Next up is special use.

G

It's the registry actually tickets for this draft.

A

Yes, no tickets, one of them is a. We need to create a registry of special use with IANA I, don't know if someone has already taken that or what the plan was for that. That's.

B

All right, that's in my draft, so I think we should just leave it there, yeah and again we'll get to that in a minute. Okay,.

A

So just just define me mmm come on excellent, so that does this mean we wait behind Barry's draft for Cola can wait synonym swear together. We.

B

Can send them through together? We can send this through ahead of time. Just have a an RFC editor note, if necessary, to say that it has to wait for the registry creation, but I don't think it's going to be an issue, because I can't see that my document is going to take any significant time to finish and.

G

I will need some time through the fixed dose came up, come on the examples, cool.

A

All right, so we wait for another draft and then I believe it's ready to go to work and go to law school once that wordings put in there yeah excellent and that's all that counts office. Thank you.

A

Next, we have the proposed additional work and we're pretty much on time. According.

G

A

The agenda so, let's start with special it's important: okay and.

B

The first bullet is me, and the second bullet is what it does. The the background is that, as I said in Singapore, the guy from Google and I wrote this up a long time ago, and it just sort of it got discussed a little bit and and sat and so I brought it back up for this working group. I've made one row of changes, so there's a zero one out there now and that covers all the comments that have been made, except for the elimination of spurious words, a spurious word in there.

B

That is queued up for the zero-zero working group version. Should this working group adopted so I asked the chairs to call for objections to taking this on and then as soon as I post, the 0-0 working group version. I will ask the chairs to working group last call it because I think we're ready with it. Yep.

A

Absolutely are there any objections to.

H

Neil Jenkins just be registry that we're creating we're going to want to reuse with, came up here and might be useful for the registry. Actually to note when, if there's something that, if the, if registrations apply to both I, don't need to just check in your FSM I mean we won't need to say in the registry. That's specific about Jim Apple I'm up I'm clean your things like in.

C

J map, you just say, you're using this registry, so it's not problem I mean unless I don't think there is any special information that you need in the register. Do.

B

H

Well, for example, in terms of differences in j-mac, there's an inbox special use because there's not it's not a fixed name, you can have a localized name, unlike I'm IMAP, so there's no current. So if you added that at special use registry, would that also apply to IMAP or would that be Jamie phone me I.

B

Will I will add a column to the registry for use restrictions or something like that? Thanks.

F

Kristen, you and the other option is to put that in the Jay Maps back in the I on a consideration of the Jay Maps back where you say where you add the you modified the registry to allow use restrictions so.

B

That's a little icky because you've in the j-man spec, you register the IV inbox, special, the inbox special use and there's nothing an IMAP that says that doesn't apply here. So I would rather have something in the registry that says that this applies to all uses of the all users of the registry unless specified here and then in the Jay map case, you can say Jay map only.

C

Yes, I'm, okay, with that um I was about to add to the bike shed by saying, would put it out in a separate document, but it's probably not necessary. Sir.

A

Excellent, so everyone's happy right, we have had no objections, Barry's going to upload a new draft which is going to go straight to work, improve last call and we'll have a registry by next week. Good work team. All right next is me the unique card, a draft which I wrote a few days ago. Just.

B

A procedural thing for just to make sure everybody understands what the registry, but the registry actually gets created when the ie, when the iesg approves the document, not when its publishes in RFC. So you actually get the registry on the soon side of things. Excellent.

A

Unique ID I wrote the first draft using names that were derived from the xgm google names. Everyone said funny idea: why don't you make them match jamup names, so I did that I've published a draft zero one which has the matches of I map names instead of Eminem's. Instead, there was debate over whether the syntax for the status response should put brackets around the value, since it is not an integer, it's not a simple value. Aleksey. It's your draft that we're messing with here.

C

For somebody who didn't get a chance to of show radio draft can give you do you have an example? Yes,.

A

C

I I think I need to double check status, might have it item as a possible value. So if you have sixty four-bit thing there, it's it's still an item, but so I would.

I

Prefer you you.

A

C

The syntax yes cool.

A

All right, so this is draft.

F

So I wanted to ask Alexia direct question about forty four sixty six. So there is a restriction in forty four 60s I raised this issue. There is a restriction in 44. Sixty six that says the arguments to status items are either a number or a parenthesized list in the syntax and I wanted to know why you did that was done and why you didn't just allow a string as well.

C

I, don't remember yeah, so I think we probably didn't have example of a string at the time.

C

Okay, I think I probably have an action item, I think in this case yeah I'm.

A

Quite happy to make it parenthesis list with an a string. You know, that's absolutely fine.

A

It's the real question. I think for me is whether there exist client implementations, which would be confused by having a bear a string rather than an a string in a list because I know, certainly with our fell client. Libraries does wound up with an extra level of indirection to either.

C

Item that was sort of the reason why there was a restriction on a regular syntax, so that client well coins are more likely to generate these things. But if the different components is trying to parse it, then it should be able to skip the stuff that doesn't understand. That's why you want to put it in it garance. But let me come back to you on this. Oh and there are on where they think it's okay, syntactically.

A

So that that's one question about the idea: the other question about the IDS, which we do not have sorry crushed, can't bring the slides back up here. There we go.

A

Again, I haven't put it up on the slide for this, but the other big question is the format of IDs. Do we allow one to 255 characters of seriously crushed again of random crap, or do we limit it to a subset of characters? Neil and I are discussing this over breakfast this morning said that URL, safe, base64 care per set seems like a sane set of characters.

A

That's a disease in capital and lowercase 0 9 and underscore that's that gives you 64 code points that you can encode your ID in allowing more efficient representation than pure hex, but it's still safe for all the contexts, JSON or URLs, in which you're likely to see one of these things. So if your client doesn't understand about escaping things properly, it will still not break anything. This is based on feedback from aren't from the list. We're planning to do exactly the same thing with IDs and in J map as well like.

C

This seems very sane.

A

Excellent I will update the text to limit it to that.

A

So work items from here for Alexi to look into the syntax for the status response for me to look into the document in the IDs.

A

So there are three things that this gives you. It gives you a mailbox ID as a status response which is supposed to, but not absolutely required to remain the same when you rename a mailbox. So in every case there isn't a potential implementation which you can do without requiring any additional data storage. You wind up with a less efficient behavior, but a rename doesn't guarantee you the same mailbox ID at the far end of every name. If the server has no way of storing an ID, no.

H

Dinkins just keys why cuz? Wouldn't in that case you just not implement this extension. If you can't actually provide any useful data again.

A

It's where you have different things supported by your server. Some mailboxes support, all this, some don't or.

H

Is where you could do unique IDs on the messages, but not the mailboxes yeah.

A

For example, or any of those cases, you can still generate a thing, but it's not well.

H

So it's nice I think this particular extension. Client waters are actually probably not pretty quickly or like a lot cuz. It's really useful and knowing what they can rely on is kind of I, don't know I, guess, I, guess and also getting something useful, but it pointless. If you say you implement it and then don't actually give any useful information from it. Yeah.

A

I used to give them an idea that they can use as a cache key.

C

Again sorry so I come throughout the draft yet, but maybe you want to split the capabilities into mailbox ideas and message: edges I, don't.

A

C

The downside of that is that.

C

If you're likely to implement one another that might be around it'll.

A

Be fairly rare, it will, it will either be all or nothing.

A

Yeah one thing in particular that I was careful not to mandate was threading because that's charge. Learning strategies is much more complex than just storing any code. If everything thank you.

G

A

Has been some debate over whether it should be nil or whether it should be a string value derived somehow from the mail message. Email ad I am in favor of nil because it gives information.

A

The argument in favor of having me having it be the same as the message ID and then you know it's the same thing: Gmail responded. They used the same ID for both we used to use an ID that was obviously derived.

A

It was the first n characters of the message ID and we changed it in our server deliberately so that they didn't look the same so that someone some client author, looking at a message, ID and a thread, ID didn't say: I- can derive one of these from the other, with some string string, manipulation- and we didn't want that. So we now XOR it against a number that I picked out of somewhere other and hard-coded in the sauce, the it doesn't matter, but it just means it doesn't look the same all right.

A

So there's there's some worker there and probably a little bit more discussion of what to do. It's ready on the list because we clearly didn't have consensus amongst the people have been discussing that so far. So I will take that back to the list as well, but.

C

You can ask whether people are happy to take it as a working group document. Yes, because I think we're so blessed passionately discussing and because we're public here yeah.

A

All right any objections to taking it sa working group document.

A

Jabba anyone give him a moment.

C

Alex he just as a participant haven't been talking about this for years, so I think that's a good! Yes.

A

This election from Jeff and they the discussion on the mailing lists in the last few days has has been people came for it. So there hasn't been anyone saying no y-you, don't miss excellent, hey, Lexie, I.

A

Have your slides I will pull them up now.

A

After I kill off the multiple copies of crashed, libreoffice that are still running here,.

C

So yes, just as a side note, it's very useful to to do slides when you already present it in this few years back. I think it was like three or four years ago when actually presented, and this one was in Absaroka group back then.

C

A

Always fullscreen full-screen mode, and it's come back over here thanks. Alright, maybe I'll just do it in mostly fool. That's.

G

A

You go sorry Linux. How does it even work math over here.

C

And okay, yeah I, don't I'm not going to go through the history of this I. Think I've been working on the document for years now and three or four years, and it has been 15 years since I'm at full of one next slide. Please alright.

A

There's a question from Java from Stefan saying a red eye map for Rev too long ago. Wonder what if anything, has changed so just.

C

Depends which version you've read but I think I'm trying to summarize all changes so far in this slides. So.

C

For the most part, this is presentation of how I view this. Obviously, if we take it into the working group, different people have different opinions, so we can tweak things, but this is where I'm coming from with this document, so three main goals. One is: we have lots of useful extensions that have been deployed since I'm up for f1 and that they make clients joke much easier, they're, relatively widespread, so let's add them to the document. At the same time, this is not a major rewrite of I'm up for f1.

C

Most of us already implement these extensions. These are just minor tweaks and there should be a section just saying you know what you need to do, update on the server to be compliant and the other thing is, you should be able to implement both imap4 for everyone under F 2, on the same port, because you use enable for Emma for f2 that serves as a switch for the server.

B

This mm-hmm excuse me this is Barry. Should we put ID I haven't read this recently? Should we roll the list, extensions and quota into here, hey.

C

Let's defer this in too late in the slide. I have a slide on this here, all right. So next slide.

C

There are three set of changes that I was planning for this document, roughly speaking additions as in adding common extensions just to the base clarifications and fixes, you know updating references and then at the end the probably the most difficult bit is taking a few things away that we think is really obsolete, so the first set of changes is I've folded. All the errata I had dated CLS recommendations.

C

Chris published one document on recommendations for MEA, and you know that effect. This document I had a separate RFC about Els or identity verification, so both of these are now referenced.

C

Then I added response, codes and pieces of I'm, a PI BNF, which I think was more relevant into the to this next slide. Now. This slide is talking about various extensions that I already added I.

C

Can talk each point where people can just treat volume or parties making protocol more symmetrical like adding an unselect edging your ID plus, because we're you know in the I'm up for f1, we have some commands, which are only using message, numbers some use, both your IDs or message numbers, and you know the way, your G+ extensions, that made it more complete. Toby's.

A

In discussion here like whether we should be doing literal rather than literal +, fine, given that that was ours today, okay gives you a nice balanced.

C

Usage, fine, then also add very common things like special use, which is very popular namespace bits of list extended. It's not fully integrated I think there is some work to do basic syntax and basic extensibility. Is there because we have so many extensions that throw and us use enable because you want to coexist with them a fateful one you search instead of search really, but the syntax more compact blah-blah-blah-blah-blah few other things. Also, our literal + I'm happy to discuss literal idle look at appendix B.

C

It should list no I'm, sorry sort of has to list one list is the changes already done and that the list is changes planned with some extensions marked, is being done, but I'll make it clear in the next revision next slide.

C

C

Future directions I would like to take this.

C

Clarify a few things like booty structure and how its generated for different my messages, I think them are for f1, had an extended example to which I have to refer to every time. I have a problem with a client or the server on this, but I think we can have more examples to actually show special cases, in particular folded message with message: RFC 802, because it's really special case every time.

C

Then there are a few key boards that I think are in common use. That I would like to add, and maybe fishing is another one and junk, but we can discuss order.

H

No drinking's is a reason they need to be there when they were in the registry now, and we just say refer to the registry.

C

Yes, the reason is that well, at least as far as I know, there are servers that don't implement extensibility, they don't implement arbitrary keywords, or there are servers that have limits.

C

The one that have limits are less problematic, but the one that don't implement. For example, Microsoft Exchange.

C

For them, I think listing the one that we really care about is important, so if they can upgrade easily, then they can it yeah.

B

This is Barry I, think in general listing the ones we think are important is useful anyway, I mean saying really. These are no longer extensions. These are things we consider part.

A

B

A

Those of you who came in like by the way we're officially five minutes over in time, but we have a half hour break before the next session. So we decided we're quite happy to keep running on I, see no, no problem with that right.

C

And then I think the one thing which is I haven't done is we do a lot of extensions to status and also status with lists so I think probably adding this to the spec would be also good idea. Next slide right, I know it's potentially at home or something that we might talk a lot about. This is things that I think we should take out if chairs want to discuss them now or not, though these things are, you know you tf7 encoding for mailbox.

C

I, think the world has moved to utf-8 to a large extent, I think we should accept utf-8 that I'm up, for that one implies that you can use them but doesn't provide enough details. I suggest we change the same. Tax service can still encode them to you gf7 internally. If they want to support IMAP for f1.

F

So on the UTF, seven mailbox names, I.

F

The way I'd lean is allow clients to express mailbox names in utf-8 in the protocol, but the server must downgrade the media to UT f7 for for backwards compatibility, we're trying to run both ports now, if, if a client does enable utf-8, they go away, they go away if it, but you know I, know that's clunky, but you're not going to break any software. If you do that well,.

C

Because this already requires enable IMAP for f2 I, think that's sort of similar to enable utf-8, and you just have utf-8 clean mailbox names and.

A

The server can store them.

C

A in modify view to f7 on.

A

The alright there's still some issues with encoding of separator. If you have a mailbox name, it has the separator encoded in it as modified UT f7. You still can't express that as utf-8 cleanly a.

C

A

C

A

This to the note, I'll I'll have a look here: I would all up for mailbox names to be a list of individuals names rather than a. This is separator, but this is ship.

C

And then I'm not trying to get him to this you at all, but there is only one through satellite, oh really, that every server should implement I remind that I agree.

C

Then the other thing is take out the recent flag, which I think most of the people find it to be useless, and it's just such a pain to implement. It requires extra local mailbox. If there are multiple clients accessing the same mailbox, free disclosure I will you know the server I'm working on just never. We just ignore this through this requirement in I'm up for that, one I think nobody complain and the other thing is.

C

There is unseen response code and mailbox open, first and seniors and authorities out I, think most clients, don't care and clients that care can issue such a command for this. So yes,.

A

And if you have a search you can ask for the first match exactly you get the same girl.

C

Yeah, you do men, he said yeah, okay, no! No! That's too much.

A

From jabbar backwards-compatibility deserves nods, not fully aware of all the implementation, headaches. Udders declare a winner. Definitely think that binary body encoding should be included in Homer forever to want to take the opportunity to have a BNF checked and, if necessary, corrected has a be enough in this draft, been given much formal analysis yet.

A

That Sean Leonard.

C

I'm not entirely sure what formal analysis is in this case, most of the things I've done is existing a BMF last pieces of I'm, a baby NFRC 44 66. It's entirely possible I've made a mistake, but this will go through three jewels, so we'll have to make sure let's try to before it.

B

And I didn't understand that I mean you rat. You appeared to run a couple of things together. I didn't understand a bit about compatibility. What he's looking for they're.

A

Definitely think binary body encoding should be included in the before that. There was stuff about countability, okay, I grated, making it clean backwards. Compatibility also deserves a nod not fully aware of all the implementation headaches to declare a winner, though that's that's. Okay,.

B

So it's a question to Sean is Alexei. His original comment that you can just you can run this in parallel with imap4 red one. Is that sufficient to deal with the compatibility for the moment.

J

Hi: okay, sorry, those were actually separate comments to separate separate parts. My point about compatible or my point about making it clean in backwards. Compatibility specifically was addressing the UTS 7 vs utf-8 discussion, because other people in the jabber list were chiming in about making it clean, so I just wanted to clarify. That's a separate thing from my other comments: okay,.

B

Yes, I recognized that it was that's why I was trying to figure out where the dividing line was cuz. The.

E

Whole thing is once.

B

Yeah I hadn't I hadn't been aware of the earlier one at night.

E

B

Okay, thanks: okay, okay.

J

No problem, yeah, the binary body encoding and the stuff I said after about 18f are totally separate. Yep I agree with.

C

The binary so binary is actually one of the following slides, so yeah, it's bad come back. Well, so we're getting backward compatibility. There is actually a short sentence with a short section which just stop sentence. They're, saying it's: okay, but I think we probably need a bit more tax, Anna and the purpose of it. We need to help out implementers to maybe have a list of all the things they need to che change if they want to start from Lima for f1 to bring it up to day to this. So right next slide.

C

So, okay, I think binary. I would personally like to add it to the protocol, and we can ask on the mailing list and I would also like to get three devel sub, because we have list extended which absorbed the whole thing. It makes it's a bit easier, but.

B

Not talking about Elsa, okay, the I would have two things. The eai utf-8 stuff should be part of the base and quota as an optional feature. I think it's better to have it in the base, rather than as an extension, but but specify it as not required to be supported. It needs to be.

C

With editor hat on, my preference is to have a section on other recommended extensions and don't pull quote Ian, because this is going to be at home. I know where, as in right, okay and and this can a the reference to holder FC where you published or whatever gets it, gets to replace it right. Yeah.

B

I'm good, if you say these, these are also recommended that that's good.

F

Okay, a couple issues I want to raise the first one is I'll note our server implements list status and will implement list rights and will not implement list extended.

F

The reason we won't implement list extended is because it allows multiple, multiple mailbox wildcards, if you, if you restrict this to single mailbox, will do list extended. I.

C

Was young and ambitious at the time I think I can't I I can live with this I think it's yeah I tell you a secret with in our, so we didn't implement it either. So.

A

Making that work.

C

You can sue dualistic standard I'm at sort of to it.

C

Okay, well, I, I think these days you can just pipeline multiple requests. If you really want you know listing box unless also three of these and everything in this namespace, so it's actually not yeah I already have to ship, so yeah, I I think that should be ready reason. You know reason both change yeah.

F

Okay, the second thing I want to consider is.

F

Either mailbox names or NFC normalized or the server can change a mailbox name when you create it.

B

E

B

I think for former: what would you what change it, how I.

F

Would change it to if it's not if it's not NFC I change it to NFC on creation? I can't do that in the I'm at bass, back because of the modified UTF, seven ward right.

B

No I absolutely that the server needs to enforce the NFC, absolutely Sean.

J

Hi I don't have a lot of experience with changing mailbox names for normalization purposes, but I would say that one way to analyze it is to look at how file systems do it and the advantages and pitfalls the consequences of what, when they do that, because some file systems don't change at all and they survive. Others like I, think apples, HFS, normalized to NF D and that has different consequences. So I just wanted to make that up. So.

F

Really I'll tell you the reason I'm asking for this is apples. File system leaks this crap all over the.

A

Place- including into.

F

You know and mailbox names and everything and add it, and then, when you take that and try to feed it into like a an index search system, you got to normalize it and it just creates this disaster so that.

A

We have unique odd days on mailboxes now and if we defined a return of the unique ID at the time you created the mailbox, you could use that to discover if anything, weird it happened to it. That way. That would be adequate for the problem. May.

B

Be but the problem is if you're using systems where you might type in a mailbox name, then you're, and you have issues with the normalization so and the reason showing from NFC rather than an FD, is simply that we've been using NFC pretty consistently in other protocols. Yeah.

J

To be clear, I think if you have to standardize on a normalization form using the NFC would be more appropriate because of the IETF way I'm fairly certain. Once you have something in one normalization form such as NFD, it should go or transmute to NFC. Okay, it just means it gets recomposed. Although there's a lot of Unicode.

B

J

That you have to deal with I mean I hope nobody's using NF KC for anything but yeah. Okay, that's I'll! Stop now.

A

However, you have created an action item for Unicode, a draft to have both append and create potentially return the ID it straight away. Interesting, yeah, yeah,.

F

That's that is really interesting. Okay,.

C

Okay, so Alex e as a participant I, don't object to the normalization of some text about this, but I need to. We need to think about the text how you returned the new name to the server I would also like to possibly fold unique ID into this back I.

C

Would like to fall to the document which is already specified, so hopefully it will be RFC. You know before we actually do this so I'll be pushing that one.

A

C

Yeah basic that the point is try to avoid too much experimentation with stuff yeah. You know we don't want to put something there which doesn't work. I want to do minimal changes yeah to there. If.

A

There's a safe way to returnest as part of the create response, then that shouldn't be too difficult, especially now that there's no encoding issues. Okay,.

F

I'm support about adding binary the proposal to fold in all of the utf-8 stuff from AI I'm, not sure the mailbox name stuff from the e aí spec. Absolutely.

F

Actually, the new EA I expect dropped all the username stuff. So so you have to use plain if you're using a non ASCII username, you have to use authenticate plain. Yes,.

I

F

Yeah, you could deprecated login.

F

No, we can't actually no I could but.

A

But it's nice to think about you can't. Can you enable before login I guess if you can enable this before login, then you could deprecated things if you're not enabling things before our syndication I can say from our proxy architecture. That would be a freaking nightmare because you don't even know which back-end you're on before he was indicated.

A

You might wind up unable de ghin once you logged in I.

C

Think there might be one or two more slides, just scroll down right. Okay, I want to get to this point and the rest we can take to the mailing list about a specific puzzle were to add on so.

B

For Chris go back to the previous thing: what I want the imap base if we're gonna Rev the I'm at base I want the I'm at base to support the ei I stuff.

B

If you have an AI message, you know the EI I downgrading and all that crap that it does I.

C

Yeah I think we need to think about. I am supportive of trying to figure out how to make it at least ready for it, but.

C

B

Bullet 2 on this slide, I'll be I'll, help you and clearly that means that I agree with the first one, although, as I said earlier, I think we should not we can we can adopt it, I, don't think we should crank it up until we get some more of the documents off the queue. Well, that means you and I can still work on it. In the meantime, right.

C

My preference is to adopt it and we just take our time and then I mean, but but not deliberately, trying to be too slow. But just you know there is no pressure and then we'll figure out it as we go all right.

A

I'm happy to call for objections to adoption at this point. Anybody.

A

Fantastic or I will adopt the work and.

A

The agenda at this point said any other business. There's nothing else, so I'm not going to bother changing back to the slides.

F

So I had published a draft a while ago, that is expired, which was the unauthenticated extension and at the time I brought it up. There was no working group and and not enough interest expressed in implementing it to get it adopted by yo. Ad I would like this draft done, because it's an it's an interim I've implemented a private extension, but it's an interoperability thing. So the purpose of this is when you have a you: have some sort of appliance or application. That's talking to an IMAP back-end on behalf of multiple users.

F

It allows that appliance to have a connection, pool and leave the connections open and switch users within the session so that you don't have to renegotiate SSL or anything like that. So that's the motivation, specific use cases of it are like a voicemail. A ploy Smail front end for IMAP that would feed voicemails in or j-mac proxy to an IMAP server would also, you know, might also be interested in a connection pool. It sounds.

A

Like it fits in one of these earlier slides we'd make a protocol more symmetrical that yeah.

F

Now it you never I, do not want it. I, don't think it belongs in I met for up to I think this is just a separate extension you implement. If you had. If you you put and I am the draft actually advises, you leave it off by default. Yeah I I haven't.

A

I haven't read the draft I assume it, it clears old, enable flags and everything else that's been done as part of the honest indicate. Yes, yes, yes,.

B

This is Barry with that edition, I'm happy with it. The.

G

Server I wrote.

B

Has severe isolation stuff in there and I would not implement it, but great I.

C

Have actually implemented it, so, although it was experimental branch, so I was never turned it on I know: I haven't that.

A

Sounds to me like a call for adoption. Are there any objections to us adopting this piece of work?

A

Excellent, we'll adopt your expired draft and start working on.

A

J

J

This is Sean Renard I'm in favor of adoption of this draft. That being said, yeah isn't there like a procedure supposed the list and give it three weeks or something like that or not? Oh, absolutely,.

C

J

C

It wasn't an Epicure, you need to double-check. Yes,.

B

I he's our ad at the moment, but might my view on on this sort of thing is that the the working group is agreeing to take a starting point? It's not a big thing and if the, if the working group later says now, we've thought about this event, we've talked about it, it's full of get rid of it. We can get rid of that. I, really, don't think we need to go through a formality of making it wait two weeks, but we will defer to our ad I.

C

Think my point is you need to ask the questions. Are there any objections on the mailing lists? Yep I'm already happy to two four chairs to optimize it adopted and there is. If there is a storm of objections on the mailing list, then it can be honored I. Think that's. Okay, cool.

A

I will I will run a call for adoption later today and and if there are no objections, we start next week, yay for all of the the four things that we've decided to adopt today, making a stack now ten documents we're working on at this point unless there's anything really urgent, I'd like to give up this chair and allow the the next session to set up, because we have run nearly half an hour over our allocated time. Thank you. Everybody.

A

C

Thank you for this.