Internet Engineering Task Force 108, 30 Jul 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF108-EXTRA-20200730-1100

Description

EXTRA meeting session at IETF108
2020/07/30 1100

https://datatracker.ietf.org/meeting/108/proceedings/

A

I'll say doing the crossword I like to be able to write on the paper I get you can do it on digital, but.

B

On my device- and I I much prefer that.

C

B

I can I carry so I have hundreds of crossword puzzles that I carry with me when I'm traveling, because they're on my ipad- and you know it's not so easy to do- that- I used to carry a clipboard with a stack of crossword puzzles clipped to it and it's just much easier to use the device.

A

B

It's just part of our saturday ritual.

A

Anyway, yep yep I'll have slides to show you in a second loaded, the slides, and what do you know? The data tracker is not showing them when I go to the agenda page unless I'm logged in so I had to switch back to the other thing at the last minute.

A

After all that, let me share my screen.

B

All right, so it's what 9 9 p.m. For you guys.

A

Yep and you should.

B

Be able to see my.

A

See my slides now, oh.

B

I see your screen ietf108 madrid virtual. Where are the topics.

A

And and here's my face awesome welcome to the extra session at ihf 108 madrid in sunny, warm and and slightly covered filled. Madrid- let's get straight down to business, so we have the no well, of course. As always, I think everybody knows this by now. You are operating at an ietf meeting under ietf rules.

A

Our agenda, as discussed um on the or sitting on the list has these four items. I had originally put ken's draft for jmap sieve on the extra meeting for some reason, but I think it probably belongs in jmap. We can deal with that in any other business. If we get there and decide that it's important to us and with that said, let me find the slides that do.

A

Exist, I'm at four rev2 and hand over to alexi.

D

Hello, can you hear me and now I don't see, slides.

A

Yeah yeah yeah, okay,.

D

Almost like I hit the wrong button because I don't know how to use meter code yeah now that all right, I'm completely understand I struggled with this yesterday.

D

Fuji uh for chairs, is slightly more difficult because you have to flip between you know, tabs or whatever, if you are sharing tabs and anyway right. um So I just uh I wrote the slides in anticipation of posting uh their 17 and then yesterday I remember that I forgot to post their 17, so I posted it yesterday, but I will cover this in the slides next slide. Please.

D

You can skip goals. This is just a reminder.

D

Right so major changes um we received, I think a private feedback directly to me and barry um about people thought that uid validity will be reset every time you expand your message so the text that was never the intent of the text, but the text wasn't entirely clear about that. So I think uh very suggested. Various digital changes to address this and I massage them a little bit.

B

Jump in for a second, um the comment specifically was that, uh because uib validity description said that um the a combination of uid validity and uid identifies a particular message forever.

B

If the message disappeared because it was refund, then that would mean that that combination of uid validity and uid would no longer be identifying that message and therefore you'd have to change the uid validity, and we initially said that. Well, no implementation has ever interpreted it that way before, but it was certainly easy enough to change some text to make it explicit so there it is.

D

Right thanks for the clarification um the next set of changes is we had a discussion about how to announce a normalized or maybe even earliest footer names. Mailbox names uh on creator name, select, examine and also append, which I forgot to mention here, um and uh I think we basically settled that there will be unsolicited list response.

D

um In this case, um there is an open issue, whether it's unconditional or whether you know there are conditions to it. uh Well, I'll get to this in one of the following: slides.

A

Yeah, I have some feedback to give on that um still to do with mentioning that, if object id is supported by the server, um it should should return mailbox id as part of any unsolicited list as well. But I will write that up for you, sorry, I haven't got to it yet.

D

Fine yeah, that should be a relatively small small magician. So that's fine, um then, uh uh in relationship to this there is a clarification that unsolicited lists are not allowed in other imap sessions.

D

If people have issues with this, please let me know, and then um I went through because we pulled in so many documents.

D

I was checking whether I ana considerations for all relevant registries were either in this document or war preference brown. If you're typing, you might want to mute yourself, sorry um so yeah there was um basically there was quite a lot of text in list extended about selection options of return options. um I decided at this point not to uh import the text and just do it by reference, but suggest that this document is all is also used as one of the reference for the registries.

D

um I think it's mostly editorial, but if people have opinions one way or the other, please let me know and next slide.

D

um Yeah, so um I actually, I know zero director. I used to do this on other people documents, but I finally decided to actually um check abnf and there were a couple of um abn from terminal missing. um Luckily they were very easy to fix, so that was that um search.

D

um There is extra text saying then body and text search keys, don't have to be to use substring search because a lot of people do more fancy searches. We had a discussion.

D

um About providing various options to do this, um I think this is uh over complicated to be honest and probably hard to implement for people. If people have a very strong opinion about this, please let me know, but I think the current text is allows flex for all current implementations.

D

um Another thing that was done is uh utf-8 was already allowed in quoted strings, and now it's allowed in responses which is uh free form, human, readable parts of responses.

D

D

D

D

I went through xml document and there were actually uh several hidden comments in the document which were there for a long time. So I thought um yeah. Not all of them are pretty, you know very significant, but it's probably worth going through them and just making decisions.

D

I think uh other than the last two issues. I think everything else is I I might not have mentioned this ever before so, but they should be easy. Hopefully next slide.

D

Right so the document has some text about compatibility with earlier versions, which I assume is I'm up to this um and then pretty much everywhere else. We rely on uid validity and uids, so I suggest so. Basically there is text saying if your id next is missing or if your id village is missing.

D

I suspect that a lot of clients actually wouldn't work. If these uh responses are missing.

A

I just need to.

D

A

At four rev two anyway before they can use I'm at four rev two behavior, so presumably at that point the server needs to return these things.

B

Yeah, I I was thinking the same thing that, because this is explicitly um enabled the compatibility is not an issue with that and in general, do we really does anybody still do imap 2 anymore.

D

Well, that's the thing you know the same text was in a map for f1 and I haven't seen any map to this implementations around for a long time. So.

B

Well, I think we can learn anything from my map too. At this point,.

D

Fine, so um I think suggestion is basically drop text, saying if they're missing and reward the first quoted sentence saying well: okay, maybe maybe I'll keep it for for reference, but maybe drop it. Okay um sounds good next slide. Please.

D

D

Multi, uh so at the moment document is so which switch from search responses to research responses. um However, the document is silent where you can send incremental search responses.

D

So, like an example at the top, this is a single response containing all matching uids, 2, 10 and 11., and at the moment the document doesn't say whether the second example below just sending two and you know, 1011 or sending 2 and 10 or whatever, but the server can actually send responses as it accumulates them and send separate um e-search responses that the client will have to then amalgamate into a single list.

D

um I think for clients. Implementation is probably it would be better to say that the latter case is not allowed unless people really want it.

D

um The example at the bottom. We actually have one other extension which allows separate research responses, but they have different correlators. They they apply to different mailboxes, so sending them incrementally. You know, I think it's quite sensible, so proposal for this uh is to say, unless specified otherwise, in an extension.

D

um One search command should result in one e-search response. Basically,.

D

D

D

A

I was wondering whether it should be should or a must for for the only one response. If it's assured, then clients need to be able to handle it anyway. So there's no point.

B

Well, I'm not, wouldn't you know it should, or must it would be. This is how it works and- and you just give it as one response- um I don't see any real benefit to a server for doing multiple responses.

B

Especially since we've gotten rid of the ability to search multiple mailboxes right.

D

It's a separate extension, so I think this is still going to be fine yeah, and this is the example. Multi-Mailbox search is a example at the bottom. That would still be allowed.

B

Yes, one one per mailbox right yeah. I think that's the way to go.

D

Makes clients slightly easier than clients don't have to expect multiple responses to deal with so all right next slide.

D

D

um Yeah not a terribly important thing, but should we say that no slash system flags are going to be defined in the future or maybe you know, maybe it doesn't matter.

A

I think this is the same thing that came up the other day with the idea that that you would create a charter for a working group. That said, this working group will recharge her later.

A

A future standard can override whatever we do here. So it is pointless.

B

Sure it theoretically can, but if we actually say that we don't intend to yeah flags, it would be a really bad thing to do, because code would now be expecting that I, what? What is the benefit for saying this.

D

um I suppose I'm trying to figure out whether people are treating slash flags separately and how they treat them if they don't recognize it, but maybe they're just.

A

Cyrus map stores in a separate bit field has 32 bits available, of which we're using about eight. Now I think, some for internal stuff and some for system flags, but if we went ahead and defined a bunch more then we'd just change the data structure and cyrus map to make more space and you'd have to do.

D

A

So it's not a big.

D

And you can still use bits for the dollar keywords if you want to.

B

D

Treat them especially so it's not.

B

Yeah, I don't really.

C

B

Benefit to saying this because it doesn't really simplify implementations. I I don't think.

D

Okay, that's fair enough! I agree.

D

Okay, no change is fine, I'll just drop. The comment next slide. Please.

D

Right uh always returning list responses on select examine, braun. You commented that you would prefer for this to always be returned.

A

Single packet anyway, so it doesn't actually cost any more bandwidth and it means you can always correlate against it. I I think just saying it's always returned, makes it much easier.

A

That's a good idea.

D

Yeah, I'm not entirely sure about single packet argument, because we already have other stuff that were required by extension, to send an extensions, but it might be okay um yeah. I don't have a strong feeling, I suppose.

A

Select is also fairly expensive and fairly rare. So it's not that big a deal if it's two packets, even.

D

All right, um fine, um if people have opinions one way or the other.

D

Let's do this change next slide. Please.

D

um I think this came from timor asking if we can actually extend abn f to allow message and body part sizes to be 64-bit in fetch. um Basically, this will only affect clients, possibly servers, don't have to use more than 32-bits.

D

They don't have to accept more than 32 bits, so they will never emit this, but clients will have to handle bigger than 32-bit values.

D

um I am a bit ambivalent about this still not entirely sure how difficult it.

D

How difficult it would be to change clients how much effect it will happen, clients, but on the plus side, this is only after. We enable imac for f2. So.

B

What was the class again, what would a server do now if it were given a mail store that had a large message in.

D

It uh I don't think the question, as you describe is quite valid in the sense of uh like my implementation cannot just work with arbitrary mail store, so we have api restrictions internally, sure that.

B

D

Be able to deliver the message so.

B

Right so when you get, if you did a fetch on a message that was larger than 32 bit in size, um because some some servers can be given an external mail store, so I don't know what they do with it, but yeah, maybe they say no, maybe they maybe they crash. Who knows.

A

Yeah undefined, I suspect, yeah, I I don't have a problem with allowing this. I think most clients are going to have probably 63 bit, but the same thing that we had for the other 64-bit discussion.

A

Clients that have the data type in it will support it.

D

That does double memory usage for big mailboxes, but potentially.

D

A

Only thing you well, that's! No! That's for the the message body part, so it adds an extra what 32 bits: four bytes of of storage for the size of the body, um but other than that. It's not going to double the amount of storage, because there's heaps of other data being stored. Sorry.

D

Yes, well double the amount of storage for size elements, yeah.

D

I mean I at the moment I don't see a problem with this for our mail store because we probably wouldn't support it initially.

D

I'm slightly worried, but so probably we should just ask for more opinions and remaining lists, just double check, but other than that, I'm okay with that cool um yeah, and I think we agreed that this is not going to affect the band. So you can you don't have to deal with injecting these messages at the moment they can only be delivered externally.

D

That's sort of a bit of a compromise for milestone that couldn't cope with that.

D

All right next slide.

D

Right so um I still have a few core comments from stefan to to fix uh most editorial missing references.

D

um We also discussed several times section about other recommended map extensions, and I have next slide about this. um The only other k uh thing is one example for body structure, because this is a source of bugs.

D

I would request people to send examples if they want their being included in the document, um and probably we should have a time limit. So if there are no no actual examples, then this is just going to be no change and then we will move on um next slide.

D

If anybody wants people.

A

If people have examples, they should probably stick them into the imap test tool um so that they actually get tested against as well as just putting them in the document. I think that's, probably more valuable. I'm also aware we're running out towards the end of the time that I allocated for this document.

A

D

This is the last significant um slide is basically which other extensions people want mentioned.

D

And I'm sort of not really advocating any of them, you know maybe construct requesting and maybe object id or the others, I'm not entirely sure. So I think we need to pull the mailing list to see what we think.

D

If somebody wants to quickly talk about this, uh that would be great.

A

We've kind of folded, fuzzy search in already, which is probably the other, the other thing um and possibly special use it. That's already folded into, isn't it so yeah. I this.

D

Is better, basically right, the things that we already folded are not on this list.

A

D

So um should I probably send separate messages on each outstanding issue and including this slide and see what people say in remaining list.

A

That sounds great.

D

All right next slide, I think that's probably it oh yeah. We still like.

D

Buddy, can you uh being married to see where he.

E

B

He's not online right now, but I will ask him later all right. That's fine!.

D

I think that's it for this one, so um I know I kept saying it last time, but I think this is really close to being done now.

A

Excellent, thank you alexia and barry for dragging it towards this point. I know it's been a huge job, so I I will. I promise you a review as well for the couple of changes I've been through since I last wrote a review of it.

D

Oh yes, um you did want to do it like one or two weeks last call, so I think you should do it on this version.

A

On 17 yep, all right I will I will do a new last call.

D

A

D

That I'll do the changes in the slides, but I think they're relatively minimal yep. Well, your call basically.

A

Yeah, I think that's good. um That gives us till mid-august.

A

Yeah great excellent, thank you. um I believe you're up again soon anyway, but um let's load up the next one, which is.

A

Quota, yes, there you go welcome back alexi.

D

Long time, no no no see uh right, update the quarter. Extension next slide.

D

Please we can skip that next.

D

uh Right changes since little one, which I presented in june.

D

um There was text saying: if you see storage or message quarter, resource types and you don't have the corresponding capabilities, they are not what you think they are. um I don't think that was probably useful um in a sense that I'm not sure what else they can be. So I took this text out. Basically, um then, set quarter now is a separate command with capability quarter set.

D

I hope this is not too confusing to be for people just I use quota set just because uh all our capability starts with quarter. So um then there was uh mine alignment with I'm up for f2. The deleted messages was changed to deleted.

D

um Usage resource, limited 63 bits added a section on imap acl and there will be a slide about this um and then I think there was text about ion considerations. I don't expect many future extensions, but now it says that if you're really going to extend.

D

Quota itself, you you'll need an itf stream rfc. If you want to just register quota resource usage, then it can be in your rsc next slide.

D

Acl, so um I did discuss this actually with can and timo, and we do have all slightly different implementations, so I think I will flag it to to the working group.

D

um My suggestion of which rights are required or not required is basically get quota, takes quarter route as a parameter, which I think the only way to know it is by doing get quarter route on an old mailbox, or you know some possibly mailbox to be created.

D

um So I suggest that it doesn't need any specific rights, because it's sort of not possible to tie that route to specific mailbox really, um but implementation to differ on this um get quarter root. Some implementations require read write if the mailbox exists, but it's uh the document also allows mailbox that not to exist so, for example, if you want to create a new submailbox or a new mailbox, an existing in one of the you know user or share folder namespaces.

D

It's still allowed to query what the quarter will apply to that mailbox.

D

um So at the moment the document says it's either r or nothing. I know it's a bit sort of weird combination, because really you don't need anything, I think, but some if, if implementations want to check that and an existing mailbox name is used, then checking all right on. It is basically mailbox. Three, that's probably okay.

D

Can do you want to talk about any of these by the way.

D

Seen him in the list.

C

Can you guys hear me now.

D

C

I don't have anything to add alexi. I think you pretty much cover everything.

D

um So I think both you and I need to change our implementations a little bit, so you think you will do this. Sorry, I don't want to put you on the spot, but I said who do.

C

I don't have a problem doing it unless uh braun has a differing opinion, since we both work on the same piece of software.

A

Yes, I'm fine by me go for it.

D

Excellent official permission from the boss minute all right. Thank you.

C

D

Yeah and then set what is requires admin rights. I think that's fine, basically.

D

Okay next slide, please.

D

um So at the moment um you can have status, deleted and started deleted storage. uh You can issue for them to calculate. You know total number of deleted messages in the mailbox or total size of them um question is.

D

Should this only be required if you support the corresponding resource type, or should this be done unconditionally? I I don't think it's particularly difficult to implement either way, but if you only say implement annotation storage uh quarter type, then it's a bit weird to to be able to do status deleted, for example. So I'm I'm slightly undecided about this.

D

Do people have an opinion.

A

I think returning nil for any type. That's requested that you don't do is a reasonable approach there.

A

So you don't fail the command, but there's no data available for it.

D

Okay, that will um create slight complications for clients if they really want this information, but I suppose that's no different from amma for everyone.

D

Well yeah, it's just like if you don't support it. This extension.

D

I'm probably okay with that, um but this is the new suggestion. I I need to incorporate this in there abnf any other comments on this.

A

What are the alternatives? You'd either return zero or you would fail it with bad command. Yeah, which is kind of um I think either of them are worse.

D

Yeah, it is all right, so I I think the suggestion is return nil. If you don't support the resource type and they only require to be supported if you support the corresponding resource type. Yes, yep. Okay, fine next slide.

D

um This is the issue I discussed in june. um Over quarter has two variants.

D

um For example,.

D

If you have one mailbox selected and then your move message to another mailbox, you can separately return information saying that your current mailbox is over soft quarter, but not by not specifying the tag, and then you can include the tag by saying your destination mailbox, the one you just copied or moved message to is over quota. um Unfortunately, there is a syntactic issue with this, because response codes are not allowed close square bracket.

D

So two ways to solve this: either drop this uh over quarter space tag variant uh or we can change the tag definition in imap4f2 to disallow square brackets. uh I don't know if anybody used them in tags, they're, sort of hard to prove one way or the other.

A

I'm really happy to remove them from tags and I'm at four rev two. Once you enable I'm at four row two, you promise not to send them in tags done.

D

So you're basically saying that over quarter space tag is still useful.

A

Well, I'm saying regardless.

D

D

Okay, all right, I can pause this question. Finally,.

A

A

Should over quote it just return the quota route? That's over quota.

D

Yeah but then it has the same issue because quadrant can it's a string and it can contains anything and then it will be allowed to contain uh closing square brackets so yeah, it's like so bad one of those things. I wish that imap genetic syntax was slightly more generic.

A

Yeah, if it was a real structured format, but we that ship is well sailed.

D

Yeah yeah a cap did it right, but then again, where is a cap?

D

Okay, so um I'll ask the two questions separately. One is format for f2 about disallowing, close bracket and saying this is possible use case and then separately. If people want to keep over quarter space tag or not next slide,.

D

Yeah, I think this is other than.

D

Basically, uh there are two small changes that I need to do. Otherwise it's done.

A

All right, so if you want to publish a new draft with those changes and then I'll, do a working group last call on that. I guess: okay, it's unreasonable, yeah.

A

A

All right, according to our notes next, is me.

A

With civ mailbox id, I uploaded this just today um a couple of times, because I missed a couple of bits. The first time. Thank you alexi for your feedback on it.

A

For those of you who haven't, read the draft, it adds a new mailbox id option to acquire, and with that you can add a parameter to file into and anywhere that you're using fcc. You can also add mailbox id.

A

If you do that, then it will try and find a mailbox that has an exactly matching mailbox id to the one you specified and if so that overrides, whatever other parameters were given, um so in particular, if you're combining with create it, will first look for the mailbox with its mailbox id. If that fails, it will then look for mailbox with the name. If that also fails, it will create a mailbox with the name and the drafts quite specifically says you can't set the mailbox id from that.

A

So it'll create the mailbox for the name, but will not give it the same mailbox id. It certainly is not guaranteed to, and that's for security. So you can't set mailbox ids, because they're normally server only created anyway, the document's there. I believe it is ready for working group last call already so I'll. Ask my co-chair to do that, since it is a draft written by me, uh it's been implemented in cyrus imap for a couple of years and we're using it in production at fast mail.

A

It's basically that I didn't get around to actually finishing here. I made the call for adoption and then didn't get around to actually finishing it, adopting it into the working group, which is why it has sat dead for a couple of years. um Please read the spec and give feedback.

A

D

Looking forward to the last call.

D

They have anything else. I think this should be just published, but you know, let's get it done.

A

Awesome uh and last item is cai. Stefan is not able to join us today and did not have any slides for us, but he did say: please do read the draft and give him feedback about whether it is on the right track or not, and he will go back and revise it based on feedback. So please read the draft. Give feedback on the list.

A

With that, we are now basically back on schedule, despite having not gone exactly to our plan so just hopping over into the data.

A

Tracker there's a group of groups.

A

A

We have milestones, adopt quite a bit done done done.

A

Civ snooze, we don't have a draft for that, yet I don't believe ken. um Let me let you in.

C

Yeah I had promised to write that draft up during our interim and I believe I did but never submitted it. So that's on me I'll get that all right. Well I'll get that done today or at least by uh tomorrow.

A

Cool I'll do a call for adoption on that um other than that. Our documents listed here are ready to be submitted. um Eai obviously will need to have work done on it, and so we should probably set a milestone for that. Anyone have any opinions about when we expect that to be done by.

A

I think realistically, probably end of the year is reasonable would be my.

D

Guess thinking about christmas yeah and I've seen the the new version posted, but I haven't read it yet so.

A

All right other than that we're basically out of documents. At this point we don't have anything else listed to work on um so question of whether we recharter or whether we go quiet for a little bit.

A

Any I guess we'll deal with that when we get there that is still set at a reasonable time. Do we have any other business in the remaining eight minutes of this meeting.

F

Yeah brian, for the minutes, I had a question. It went past me. What was it that you wanted to create a new milestone for the end of the year.

A

uh Yes, so create a new milestone for submit eai uh to isg, just to use all the acronyms uh for december 2020.

A

And I will update the other milestones to be uh the times that we have already agreed for those documents during this call. Oh.

D

Cvi is a personal draft, but it's not.

A

I will check whether I've done a call for adoption. If not, I will do that call for adoption and I'll update the call for adoption milestone to be.

A

A

Sounds good all right, uh I will see many of you over in the jmap call in a few minutes. Thank you for your time, see you there.

A