XSF Discussion - 2017-11-07

no its pretty much what we already have. as long as the characters are shown i think its good. KISS.

escape characters make it overly complex

I always want to say something is KISS but I wasn't sure if it translated or might insult someone :)

Keep It Simple Stupid

thats a very old one.

*nods* sounds good to me.

i was putting together a media attachements xep last winter

i think we even added support for it

https://pastebin.ca/3905635

what we didnt resolve is ensuring that <thumb> accurately reflected the media being shared.

URLs are difficult regardless. you certainly cant have clients autoloading them because they can unmask anonymity

SamWhited, I honestly wonder if you are ignoring all feedback from the previous XHTML-IM thread. but then it's time to sleep, I'll try to post on the list tomorrow

I was wondering if everyone was ignoring my responses to all feedback from the previous thread.

:)

But maybe we shouldn't get into that, because that's going to be an unproductive discussion. Please explain what you think was ignored instead of making inflamatory statements.

Sure, I'll post on the list, going to bed now

I'm really not interested in arguments that xhtml-im can be hypothetically implemented in a secure manner. its been in use for a long, long time and I'm not aware of a single secure implementation of it.

the thin benefits it provides to users cannot possible outweigh the risks

Xhtml-im's main benefit is it's easy

That's also it's main downfall, easy to shoot yourself in the foot

we had a presentation at ctrlh two weeks ago on "sanitized" html

different browsers parse slightly different, and in those differences, you can sneak script through

in one of the examples he used a utf8 quote inside <img width=> to get the sanitizer to believe it was all part of the value, but then include a ton of extra stuff

that's a good one I haven't seen before

it was one of the trickier ones, and they all were limited to certain browsers

the other presentation was on a USB charging cable a guy had made with a GSM modem and SIM card slot hidden in it, and it would start uploading data from the users phone when they plugged it in

Arc: Would that still work after some round trips through an XML parser?

idk, i guess it depends on the exploit

some of them had quotes within quoted values

at least one of them relied on the sanitizer modifying the value to remove escapes

anyway they were all web-based examples and the main takeaway was if you need to allow html, put it in an iframe loaded from outside your trusted domain. such as every message from the same user from <user>.usermsg.mydomain so if they do manage to run javascript, it won't be able to access anything else

thats actually not a solution i had considered in dealing with the candy problem

ive been trying to find the video of the presentation but i dont think its been uploaded yet

Arc: plain old CSP header doesn't protect you?

afaik that only works with an iframe

which means every msg must be in an iframe

CSP could work, I think, if you say that no inline script whatsoever is allowed and all scripts you need for operation come via script tags

(and then some strict policy where those scripts can come from etc.)

About the proposed styling XEP: What do you need service discovery for? I thought a key point of this spec was, that it doesn't matter whether the receiver supports this XEP or not.

Syndace, yes I agree. I was meaning to propose getting rid of that. but it doesn't really hurt either so and there were more important things to talk about

Syndace, even if disco is not required by the protocol, it's still nice that you can use it to get an idea how widespread it is implemented

Flow, but it's a MUST

Syndace, and that is problematic because? (Although I can see that a SHOULD may be ok too)

I just personally don't see how the discovery should be used. My $client won't behave any different if it detects support or not. It won't strip out asterisks just because the reveicing person does not advertise support.

IMO it's always a good idea to keep specifications as thin as possible, only including things that are needed for it to work.

And disco is definitely not required for it to work

I mean, all clients available probably support XEP-0030 so its no additional work in that specific case but still, why require something you don't need.

Though I'm fine with looking at more important things first, as danial said

Syndace, as I said, it allows you to get an impression how widespread the xep is implemented

You may argue that it's more overhead for little benefit, but I think I would disagree with that

if it's just for "an impression for how widespread it's implemented" then it should go away

that's what looking at code is for, it's not like there are so many clients you can't tell

Discussions like these are why we can't have nice things, guys. Lets fine-tune later.

Guus: for example? :)

zinid: we (myself included) are bikeshedding to much, in my opinion. That causes us to not be very productive, which in turn fuels the argument that XMPP is outdated / obsolete.

I feel that we need to build momentum.

rushing crappy specs to production is a problem at the other end of the spectrum though

isn't all this xhtml/mardown rantings a bikeshedding? :)

moparisthebest, I'd love our problem to be that we design and implement things too quickly.

zinid, Much of it, yes.

like this is the most serious problem in xmpp?

where are the priorities?

we don't have avatars working goddamit

moparisthebest: now that I have you: could you enlighten me on a SSHL confusion that I have

did the config file syntax change between 17 and 18, or am I missing a file?

it didn't change, but the sni/alpn stuff was added in 18

and you used to be able to get away without a config file and specify everything as command line args, but with alpn/sni you need the config file

aaah! Then that's likely it.

I need to write up the page on the wiki again, I just lost everything and haven't got around to it

I think I got confused by the duplicate file name of /etc/default/sslh.conf and a missing file that likely should go in /etc/sslh ?

normally it's /etc/default/sslh (which is a shell script) and /etc/sslh.conf or /etc/sslh/sslh.conf which is the config file

not sure how debian packages it these days though

Guus, I don't think that this is the reason we are not productive

Guus, until I get around to it there is a bit of an explanation here: https://wiki.debian.org/InstallingProsody#XMPP_over_HTTPS

even though the section title is totally wrong, just ignore that :)

Thanks

Flow: every bit helps.

The reason is that we try to reach consensus on an to early stage. It should be normal to have multiple (partly) competiting experimental XEPs

It would probably help to get the MUC-NG thing that is currently restricted to just what MIXs experiments with

We should write a new file transfer protocol, too. Been a while since we did one of those.

also how about SRV records for connecting over QUIC

Bittorrent is hot, I hear.

Kev, Not obscure enough for us. Maybe IPFS over XMPP?

Can't we just add our chat messages to the blockchain and call it a day?

dwd: I'm referring to a particular summit discussion. 2007, maybe.

Kev: pics or it didn't happen.

It's a discussion that would have been better to not have happened.

So I'm ok with that.

:)

Kev, Ah... I think that might be the summit before I started doing anything serious.

You've been serious for 10 years now? auch!

Guus, Well, serious with XMPP is around 2004, actually. But I don't think I turned up to Summits straight away. Serious with Open Standards more generally goes back to 1997 or so.

serious dwd is serious since 1997

Well, when I say "serious"...

https://events.ccc.de/2017/11/04/people-34c3/ btw

mathieui: the people I talked to thus far aren't really interested in an actual assembly. (=having a physical space). I just submitted my xmpp talk to the fsfe assembly

well, saying "let’s meet up at time X" should be good enough

oh, that's to bad. There's plenty of you going

What is the accepted behavior for candidates in the vote on themselves? abstain once and vote 4 others? Vote 5 others? Not vote at all?

Just vote for yourself. Who cares

As far as I understand, you're allowed to vote for yourself, so if you think you should be in, vote

Except if you believe you aren't suited for the job of course

isn't that pretty awkward to candidate for a position you think you are not well-suited for?

That's the trump way

that's a candidacy everyone else thinks you are not well-suited for.

Okay, might be me for Council as well

Ge0rG, We have had people stand before specifically to make an election contested.

dwd: wow. That's pretty crazy

Ge0rG, Presumably they felt they could do the job if selected, but I could see they might vote against themselves.

Or might have voted. It was some time ago.

if I understand correctly I have a flamewar to catchup

I tried to revive this page if anyone is interested https://wiki.xmpp.org/web/Tech_pages/XEP-0368

found this paper, has it been discussed already? https://eprint.iacr.org/2017/666.pdf

RE Styling, the thing I haven't gotten feedback on so far is the type of styling itself. Does this encompase peoples needs as far as IM goes? (I know some people want a full formatting and layout engine, but that's out of scope)

Or is it too much even? Eg. is there any point in having ~strikeout~ which I doubt anyone will ever use?

SamWhited: I

oops

I must admit I've not read it in detail, but as far as I'm concerned, a basic set (bold, underline, italic, with perhaps a couple of others) is more than enough for a boatload of usecases, therefor making it a practical XEP

There's likely other use-cases, but there's no need for one XEP to rule them all, right?

SamWhited, under the assumption that proper layouting is not possible, I think the things you include there are sufficient.

Guus: is underline a requirement for you? Having that makes more sense to me than strikeout, but I don't have it right now

I don’t think that underline is a good idea in general.

yah, I'm not sure how useful that one is either; just slightly more useful than centerline

(maybe)

I’d say the opposite.

we already have two ways to emphasise things in that protoxep

That's fair too, semantically they're probably the same

Now I'm even less sure :)

I think _underline_ is supported in gajim though

yep, that underlined it

that some client supports it doesn’t make it a good thing

why not, specify what's already implemented most places, make them all optional, done

you keep repeating "most places/clients"

> We are not typing on typewriters any more. We are using computers. Word processors, HTML, CSS. Underlining means a hyperlink. Period. If you want to emphasize something, use bold, italics, indents, all caps, or any combination thereof.

https://writers.stackexchange.com/a/4680 kinda funny

SamWhited: underline is not a requirement for me, but it's somewhat of a default featureset, right?

but, to be honest, I don't care to much. My point was: keeping it simple is good.

I don’t think you can do that anymore in e.g. Markown or rST

Guus: I'm not sure; Slack and Whatsapp don't support it.

And I think they count as "most things" more than Gajim does.

I'll leave it off for now for the sake of simplicity; we can add something later if it's a problem.

Might take strikeout out too

SamWhited, what about disco

SamWhited: as I said: I'm not bothered by the exact set. I'm just gratefull you're keeping it short/simplish.

moparisthebest: oh yah, I saw that discussion. I'm fine with removing it.

Doesn't matter to me either way

I was going to wait to see if it was accepted or not before pushing more changes

> Might take strikeout out too I think the ~ character is rare enough that it won't cause trouble and slack and whatsapp are doing it. So in the interest of making transports work I'd just leave it in.

Let me tell you about my ~/stuff

that's fine you didn't put it twice

wait is ~/stuff/~ a valid path?

Zash: that's lacking the closing keyword though

I don't think I want to know...

moparisthebest, it is

why shouldn't it be a valid path?

dangerous to delete :)

jonasw, just upvoted https://writers.stackexchange.com/a/4680, nobody needs underline if you don't write on a board or on a paper

Guus, could you cancel the second-newest build please? https://hub.docker.com/r/xmppxsf/xeps/builds/

I may be stupid, but (re styling): > Matches MUST NOT extend over newlines and MUST start with a whitespace character or be the beginning of the line or the byte stream. how does that match with the example "*emphasized*foo*emphasized*" where "foo" is not emphasized?

cancelled

Guus, thanks -- even though I don’t see it in the UI yet.

good eye, that's broken

I have those rules tweaked a bit based on feedback locally anyways

SamWhited, re standards@: > A 100% solution with no false positives would be great, but I > can't think of a way to do it without significantly more complexity or a > formal markup language (which is against the requirements I drew up > based on previous discussions). what’s wrong with a formal markup language if we’re gonna need a parser which can handle context-free grammars anyways?

I'd love to know what matrix is doing and eventually try to convince them to use the same syntax (to make transports) but matrix is impossible to Google

(now I _really_ cancelled it)

jonasw: duplicate content, even more complexity, etc.

IIRC from my compiler construction lectures you need an LR parser for a CFG which is already quite a complex beast.

It's really not

I don't think you need an LR parser here though

would have to take a deeper look. In any case, writing down the grammar formally is probably a good idea

I think we're over thinking this. I've got an implementation which I'll push somewhere soon and it's relatively straight forward.

I'm not even bothering to build an AST (though doing that would make things much faster)

Guus, yay :-)

What's that, just s/\*.*\*/<b>&</b>/ and stick it in innerHTML?

Good idea

Zash, you forgot the \b before the \*, they must be on word boundaries (essentially)

Actually, if I remove escaping and strike out it may be possible to make a few minor tweaks and make this regular

That would simplify things a lot

hm, doesn’t the precedence of block over monospace over other inline already make this non-regular?

ah yah, you're right, having block level things probably still make that impossible

Just for reference. The last time I tried to escape on slack, I have given up after three futile correction attempts.

That's probably a good reason to remove the escape chars

Maybe there are no escape chars on slack, and this is why I failed. We'll never know.

I did look at their docs and I don't remember any mention of a way to escape formatting chars

At this point, let's publish something and start experimenting with implementations

Yeah. You can't escape slack

Guus: There are at least two implementations already :) (Daniel and I both have one)

awesome. This is in Conversations

/

?

Foo *bar*

Guus: https://github.com/siacs/Conversations/tree/styling

It's just an experimental branch

ah ok :)

I don't think it's complete either. I also have one in library form that I will push somewhere "soon"

Cool. Let's start gaining experience with this, and discuss it from there.

Guus, SamWhited: the current beta has this.

I'm sure it will freak out admins, developers and mathematicians

let's see if I can hack this in Spark real quick

test _foo_ *bar* ~foobar~

why's ~this~ bold and strikethrough?

ah, doesn't reset

okay. Split by whitespace, checking first and last character only - Not what's in the xep, but it'll give me some feelign on how this looks

so far, I'm not unhappy.

I've not looked at the protoXEP yet, been busy. Is it basically *bold* /italics/ _underline_ ?

*bold*, _italics_, and ~strikethrough~

(although I was thinking of removing ~strikethrough~)

Can we have /italics/ and _underline_ please? :)

If we do that, I have an implementation that predates the protoXEP by about 15 years.

/italics/ seems nice but also likely to conflict with paths and URLs, Slack and Whatsapp both do _italics_ so I felt it was best to immitate what they were doing, but I coudl be convinced otherwise

Slack and Whatsapp both annoy me for getting this wrong :)

I'd rather follow convention unless there's a compelling reason not to

(but I do agree that I think those make more sense)

See above - I've got about 15 years of convention for you :D

But it depends what you want.

but silicon valley knows better

If what you want is an input format, it doesn't matter much.

If what you want is something that falls back to plain text gracefully, then /italic/ and _underline_ are better.

Kev, Gajim also does *bold* /italics/ and _underline_, but no ~strikethrough~.

/italics/ is sooo 1997

I've not seen that since.

Guus, Yes, and your point?

who said I had a point?

I'm not sure that they are better though, regardless of how nice they look people are used to Slack and Whatsapp I think. Then again, I'm not sure it really matters all that much either way.

SamWhited: Except Slack never renders this.

If you send *something* in Slack, you only ever see the bold, not the surrounding *

So people aren't used to seeing _somethingitalic_

Oh yah, it's a tiny bit different, but as far as the characters that people type goes they're used to writing _italic_

(I like how this is rendering for me now :P )

I assume… I haven't used Slack except very briefly, I just read their docs.

I'm used to writing /this/ in Slack and wondering why they don't support italics.

You mean <i>hello</i>path/ is what you see in places ✏

This is not a hill for me to die on, but if we've got a 15 year old convention in existing clients, it seems weird to break that.

yah, I would say this is entirely bike sheddy, except that I do think the /path/to/foo is potentially an actual problem

What clients do this convention?

Gajim? Any others?

Kev, I have to admit that although /italics/ is far more natural to me, I never use it (or _underline_ either).

I put it in Psi about 15 years ago.

it's not a problem if you leave all the characters always though SamWhited

That's also fair

I'm not saying that this should necessarily be a trendy popularity contest, but I suspect more people use Whatsapp and Slack than Psi and Gajim, so if it's just a matter of "people are used to this, it's a convention" then I'd still want to stick with what's in the protoxep now.

people are used to the input convention, not to the wire format (they don’t care about it, but we should)

finally the build finished: https://xmpp.org/extensions/inbox/markup.html

jonasw: what is the number in span start/end?

it’s explained.

(codepoint in XML character data)

jonasw, Read that already when I noticed the github notification. I'm undecided on that approach. I do think some of the failure cases with overlapping spans and blocks and so on are going to be weird.

ah sorry, I was just looking for that earlier and didn't see it

What happens if I put it in the middle of a grapheme cluster composed of multiple code points?

SamWhited, it gets split when rendering.

I presume

SamWhited, We've used a similar approach in <references/>, BTW.

*nods*

depends on your rendering engine I think

I wondered about it in references too actually

I was already wondering whether that is something which can be used as an exploit, but aside from splitting emojis and possibly cutting of accents from letters I don’t see how this could be abused

but then again, unicode is weird.

jonasw: looks nice

good job on formulating an alternative, jonasw

jonasw: the introduction seems weird. I'm missing a "this approach thrives to solve the mentioned issues in the following way: ..."

on first glance, I'm inclined to prefer the less complex version, but I love to see a documented alternative

vanitasvitae, I always forget that :)

minor nit: if you're sending an encrypted body this leaks information about the body. Probably not in any significant way, but it seems worth mentioning in the security considerations or somewhere.

Guus, I like the simplicity of the *input format* Sam suggests. I don’t like having it as the wire foramt.

jonasw, understood

(also, should that have been bold?)

no, my client can only into XHTML-IM and I’m too lazy to do that

SamWhited: I think the biggest design failure of omemo is to only encrypt the body...

> minor nit: if you're sending an encrypted body this leaks information about the body. Probably not in any significant way, but it seems worth mentioning in the security considerations or somewhere. This is a problem with all the references and annotations XEPs

SamWhited, good point. Admittedly, I was in a bit of a hurry because I only thought of this last night and I felt it’d be good to have the idea out before next council meeting.

But as Ge0rG said this is rather a problem with omemo than with those xeps

daniel: I think the reverse is actually true.

This is a problem with any encryption scheme that only encrypts the body.

Ah, which you then said :)

isn’t that what daniel said?

As I've previously expressed, I think it only makes sense for OMEMO (or anything really) to encrypt text nodes, not actual XML (now you've got a weird mixed-mode stream and that's going to lead to worse security issues later)

right

We need to encrypt the meta data finally. That's what the services are after!

My hotel wifi here is *not* fast.

So even if it encrypted more than the body, I don't think it could encrypt this.

But that's a discussion for another time.

I'm off. goodnight everyone.

Night.

o/

nighty Guus

> As I've previously expressed, I think it only makes sense for OMEMO (or anything really) to encrypt text nodes, not actual XML (now you've got a weird mixed-mode stream and that's going to lead to worse security issues later) It's a quite complicated problem. That's probably why nobody has solved it yet

Mixing data and meta data is known to fail security since Captain Crunch. We should have learned from that by now.

jonasw: fantastic proposal, I really love it, thanks a lot for that !

goffi, you’re welcome

jonasw: we've already discussed this a little bit I think, but what do you see as the benefits of doing it this way?

it would be quite neat to do hop-based encryption, like encrypt entire thing so your server knows to send it to otherserver.net but not who@otherserver.net, and that server can see who to send it to but not who it came from or whatever

SamWhited, I’m going to head to bed in a few minutes, so I don’t think we can discuss this at length now.

that's getting more onion-routing-y though and is starting to sound like a protocol not-xmpp

sounds good

moparisthebest: that fails for presence

moparisthebest: there are protocols not xmpp that do it, like tox

well tox is p2p without any hops right?

no servers anyway, I think

SamWhited, it provides the protocol break we need to make rich text safer (compared to XHTML-IM) and provides the formality which is needed to make it future-proofer (compared to ad-hoc text formats).

moparisthebest: p2p in the real world has servers, only they call them super nodes or something

SamWhited, I think that we shouldn’t be mixing input conventions with wire-level markup.

jonasw: why not?

this is somewhere between a matter of design principle and concern about interoperability.

because input conventions only matter for humans, and may differ between different user interfaces

I don't see why interoperability would be better or worse with an XML based wire format vs. a plain-text-with-magic-characters based wire format

I'm saying like xmpp setup is like A (client) -> B (server) -> C (server) -> D(client)

B only needs to know it's coming from A and going to C, not about D, C only needs to know it's coming from B and going to D, not about A

I'm actually worried that making it more extensible and "future proof" as you put it will be a bad thing here. It's not necessary and I suspect will just lead to feature bloat in the long run

SamWhited, when you start to define new meta-characters or obsolete existing ones, the messages are interpreted differently.

and any content besides those specific routing instructions can be encrypted e2e so only A and D can see them

jonasw: that doesn't bother me at all

SamWhited, I know.

that’s what worries me.

that's already what happens when I add a new smiley or something, and it's a minor anoyance at worst

It's a consistent annoyance.

True, but still a minor one.

I don’t see why we have to buy that minor annoyance which is simply going to happen (I don’t approve of messengers which replace text with emoticons either, by the way) if there are alternatives.

fwiw i do see the benefit of jonasw approach. but i don't see me implementing this anytime soon due to the omemo problem

If there were good alternatives I'd agree, but I think the trade offs on the alternatives are worse.

And if it weren't such a minor problem that doesn't affect most people

daniel: fix omemo then...

SamWhited, let’s turn this around, what’s your problem with my approach?

Ge0rG, i do make some money with Conversations. but unfortunatly not enough to go down *that* rabbit hole

jonasw: It's a lot more complexity. How does it interact with message editing, for example?

How does it interact with OMEMO, OTR, etc? Does it leak information?

SamWhited, do you mean message correction?

yah, that one

message correction is clear on that one: the new <message/> replaces the old

so you’d include the markup again.

SamWhited: the corrected message needs to have its own markup xml

ah yah, that one actually makes sense I suppose

SamWhited, interaction with OMEMO and OTR etc. is at least better than with XHTML-IM (instead of leaking the full plaintext (hi pidgin!) or having no markup at all, you can leak only the markup or have no markup at all. that’s a step forward or so...)

Ge0rG, also my last attempts in starting a discussion on stanza content encryption didn't go anywhere

SamWhited: also you can just strip the markup and still have human readable body

Ge0rG: You can do that in either case

And screen reader readable

or, with SamWhited 's approach, you can do nothing and still have a human readable body

And no false positives

moparisthebest: for a limited subset of "human readable"

daniel, can you point me to it? I can’t find it in my local archives.

(a subject line or keyword to search for will do)

works for everything else so

Anyways, I'm not saying that this is bad, just more complex and I don't see any significant benefit that's not outweighed by the disadvantages.

I might have just missed it but is &gt; 1 index/character with respect to start/end or 4 ?

moparisthebest, 1.

"XML character data" => entities are already decoded

sane XML parsers won’t hand you entities anyways

But it seems that different users apply different weights to different disadvantages

I saw XML character data, I didn't know what that meant though

moparisthebest, in short, that what your XML library hands you when you ask for an elements text.

the main disadvantage is all clients have to implement this or any formatting is lost

wheras if I say something *really* important

you get it regardless

jonasw, mail is in the 'omemo discussion summary 2017' thread and my particular mail is from june 22nd

moparisthebest, that’s the better case I think

moparisthebest, with Message Styling, all *sending* clients have to implement it or their message may be incorrectly and unexpectedly formatted

no, sending clients don't have to implement it at all

if we’re going to have incompatibilities between markup interpretation, I prefer them to be vanishing markup instead of markup which apperas out of nothing

no one is using a WYSIWYG editor to bold something

they are just typing *bold*

moparisthebest, they do, if they don’t want their line which starts with "> " to be interpreted as blockquote.

moparisthebest, you’re assuming that all clients are operated by humans.

I feel like a broken record here, but again, everything has been doing it since way before xmpp

so people are used to it, and it's fine

anyways

as promised, heading to bed

email for example, IRC for another

moparisthebest: and the sending client can convert that to bold in the input line and send well defined wire format. Or you can press undo and send ** instead

Yah, it seems like it's been firmly in the "good enough" category for many, many years.

hmm then I have to care though

i can see myself adding jonasw markup to outgoing unencrypted messages on top of SamWhited's styling

it doesn't hurt

Tobias: IRT xep0385, I have an objection to table 2. First and most notable, Opus is an IETF standard for audio and was left out entirely.

I do immensly prefer jonasw's solution as opposed to duplicating the text in a different element

I agree with that

in fact, add all the xhtml-im features to that and publish

I do not agree with that and see it as the most significant downside to the spec

it feels like a mandatory whitelist kind of

I suppose mr. evil could still include <script> directly in the body

but I guess that's always a danger for a dev dumb enough to .innerHTML it

And then a client needs to put styling into the body, markup into a related xml and add xhtmlim

I don't think so

moparisthebest, what you type as your input doesn't have to be what the client sends verbatim. If you want to use markdown (or whatever else, poezio uses ^C<key>) as your input format, fine. But don't send that on the wire and let the receiving client decide on what/how to display it

pep., I get it, and totally agree for complicated formats, but don't overcomplicate something simple that has existed since before xmpp

How does that change anything to what I just said above. As a client I still want to decide what I display

pep., do you understand that I meant to emphasize *this* in my message though?

"it's always been how we've done and it works fine". I don't have an example offhand but I'm sure this is not the best argument out there

or do you need some fancy wire protocol to understand that?

for the limited subset of markup in that one particular xep, it works perfectly fine

Still, what if I don't want to understand it, what if I can't understand it

if you want to get more complicated, do so in some other way

As it's been said on the list already

then I guess tough?

at least you had a better chance than if you didn't get ** because your client didn't implement fancy-new-xep

Nice for an eXtensible protocol

SamWhited, Example 8, isn't the nested version missing a space?

pep. no, I need to clarify that. If you parse the outer quote as a block, then parse the inner quote then it's the "start of the byte stream"

Or maybe the sentence above can be reformulated a bit ✏

k

at least, that's how it ended up working in my little parser which recursively parses inside blocks, and it seems sane

I was debating if I needed to be strict about specifying that though or if that's just going to confuse people. Overspecifying is sometimes as bad as underspecifying.

But then some clients might render "> > " different from ">> "

SamWhited: the XEP could probably use a few more complicated examples at some point. Like something people could copy past into their unit tests

daniel: agreed. I started on a unit test suite as well which I'll push when I push out my implementation

But we should nail down the syntax and test everything a little bit more before going through the troubles

Indeed

Conveniently, the commonmark parsing rules are basically the same (I mostly stole them), so you can test things (albeit with slightly different characters) using their tools more or less

http://spec.commonmark.org/dingus/?text=no%20format%0A%0A%3E%20%60%60%60%0A%3E%20Preformatted%20text%20to%20the%20end%20of%0A%3E%20the%20block%20quote%0A%60%60%60ignored%20(not%20sure%20about%20adding%20this%20to%20the%20XEP%20or%20not)%0A%3E%20preformatted%20to%20the%20end%20of%20the%20body%20(not%20a%20block%20quote)

pep., point is people have typed like this since before xmpp, type like this now, and will type like this later, whether someone standardizes the rules or not

what exactly is the problem with it? not enough XML?

I think I'm going to move the block/span stuff into implementation details and make that non-normative. If a few clients have incompatible edge cases it's no worse than today (where lots of clients effectively implement this slightly differently), then I can still provide guidance without overly limiting people. </thinking outloud>

moparisthebest, you pollute the plaintext. As people have said already: it's not easily interoperable. Once you say italic is not // not __ anymore, what happens to __? people will still continue to display both because "tough luck"? (as you say) Not easily extensible. Adding more and more meta-characters removes choices from the selection of characters I can use to type text. Maybe for you it is "human readable", and for others in your own circles, but that's the extent of it. You are not taking into account other people, bots, you are not taking accessibility into account.

"not easily extensible" is a feature.

So you're fine cluttering client markups until the end of times

yes, although I wouldn't phrase it that way

right you don't want it extendable at all

it's a de-facto standard anyway

no it's not

I was reading just above Kev uses // for italics

You are using __

this happens and will happen regardless of if we standardize

moparisthebest, no that's the point

yea, I don't think it matters, I also use /italics/

if we standardize it shouldn't happen

gajim supports that by the way

That's the whole point

pep. I think he meant that most things are already using some kind of indicator in the text and people will keep doing that. Not the specific indicators

yes

I mean most people write like that anyway, regardless of any client support

and then also lots of xmpp and non-xmpp clients mark it up based on those

SamWhited, agreed, people have been using that, and as I said above I wouldn't mind the client using it as input, but I really not keen on sending that on the wire

that's fine for something that isn't already widely used I guess

My IRC client, one of my XMPP clients, my email, and probably other things I'm forgetting already send that on the wire and it works pretty well.

and as you said slack and whatscrap

No, it works for you and your circles, because they know

I thought whatscrap was for the mom's of the world

It also works for every non-technical person who uses whatsapp that I know (and that's a lot more than the technical people in my XMPP circles)

Slack people don't know, Mattermost people don't know

They only have to click

normal people don't use IRC so I think I can safely discard that one. Or they use some fancy client that handles that for them, still you rarely see non-technically people willingly use it, Same for email actually, it's not the same usage at all between technical and non-technical people

Non-technical people often send html and there you have rich content

(and fancy colors, eww)

are you saying XMPP has more users than email and IRC ?

Non technical people don't use the markup characters in normal use, because they don't need to paste directory paths nor math formulas

because uh, seems wrong

moparisthebest, not "more"

Different kind of users I believe, or at least aiui that's a target for the XEP

Because Slack and Whatsapp are cited a number of times

what kind of users are you trying to target who's needs aren't met? I'm confused

because you mentioned IRC which is more technical and whatsapp which is the exact opposite

both implement this same thing

No I'm just trying to extrapolate who you are thinking the XEP is targetting

I didn't mention IRC, Sam did

everyone on the internet does this type of thing

‎[04:48:57 PM] ‎pep.‎: normal people don't use IRC so I think I can safely discard that one.

Yes, read a sentence above now

SamWhited> My IRC client, one of my XMPP clients, my email, and probably other things I'm forgetting already send that on the wire and it works pretty well.

I've never heard of an IRC client with a WYSIWYG editor

they don't even have passwords without messaging a bot

maybe you don't know about sasl, but that's how of the scope for tonight

that's a super recent extension not everything implements though :)

and server-side, it still gets passed through to the nickserv bot

This went really off the rails in the time it took me to discuss something at work and then come back :)

I'll be interested to read the backlog and see how it got to SASL

bottom line is if it's good enough for email, IRC, and whatsapp, then it's fair to say it's good enough for xmpp

SamWhited, same here :/

because that covers the gamut of *all* internet users

SamWhited: No, it wouldn't.

No, you won't, rather.

I'd love to hear about the internet user that doesn't use email or whatsapp

You might think you will, but it'll be a big disappointment when you get there.

Kev, :)

drat

OTOH, I just found a presence leak in 6121, so that's nice.

oh fun

Actually, it's not a presence leak, just an inconsistency. Drat.

SamWhited, otherwise, if you have time to answer my direct reply to your comments that'd be nice

i could live with both approaches, if Sams approach is *really* simple, so nothing fancy like lists and stuff, keep it simple and readable even if a client doesnt support it

pep. direct reply?

like the small block of text before moparisthebest barged in

hey I've been here the whole time :P

oops :P was sent as text and rendered as a smiley, need a xep and new wire format

pep. I don't see it

jonasw, approach i like because i dont need to write parsers for that, seems simple and its more of a replacement for xhtml

SamWhited, https://bpaste.net/raw/14760ce8f04d

lovetox, completely agree there

Please hold, asking a friend who knows nothing about computers "How do you make text bold in Whatsapp?" to see what they say

SamWhited, yeah, one example won't discard all of this. I do admit I don't have statistics on this, but neither do you ✏

Sams approach also seems like a: Many clients do this anyway so lets make a XEP out of it

yea, why not?

because sometime it's not because a lot of persons are doing something that it's good ?

i dont really feel that this needs a xep, a client can show this that way and have a config option to turn it off like gajim has for proably a decade

lovetox: I more or less agree with you, but I was told that we can't deprecate XHTML-IM until there's an alternative (and maybe if nothing else this will introduce a little consistency or broaden support for simple formatting)

pep. in that case if you're allowed to argue with no data and I'm not allowed to at least ask around, I'm not sure how you want me to respond to that. I don't think it's a problem, people use whatsapp, it does this. I don't think it has a bold button on Android (unless they've added one), and people still send bold text.

The people who really, really won't understand this aren't likely to use it, so no harm done. Unless you can demonstrate somewhere with a high rate of false positives where it would cause confusion?

Or, alternatively, people who really won't understand this can use a client with a bold button.

re: no data, fair enough.

Then I don't understand why you absolutely want to pass this on the wire

i dont feel strong as other people about your approach because i simply think it will change nothing

clients use this formatting already, and will conitnue

pep. because it's less complicated than adding an entirely new wire format, does what clients already do and what (I suspect, with limited data as you pointed out) people are already used to

It also follows a convention that's been in use for a long time across multiple types of media

i think people will likely accept it more if its a : lets write down the status quo

but if you mean, lets extend it later and add all kinds of stuff

then i think jonas approach is better

I specifically want it to *not* be extended later with all kinds of stuff.

SamWhited, why?

pep. because that's the trap we always fall into: make everything as complicated and as extensible as possible. This leads to incompatibilities between clients, makes things harder to implement, etc.

pep. because you get more and more styling elements into the plaintext

I want styleing, not layout

*styling, even

SamWhited, I am not willing to sacrifice <body> over a "I don't want to complicate things", I don't think that's something we can come back from

pep. it's already something people are doing, you're not changing or sacrificing anything.

It works today, it's simple, people are used to it. That's "good enough" as far as I'm concerned.

jonasw interesting approach :)

I definitely prefer it to Styling as well. I find the whole thing futile in the first place, but I understand the security concerns and I think jonasw's proposal covers it

SamWhited: I think jonasw's approach is overcomplicated, however, it has some interesting side effects.

im trying to remember it, but there was a XEP published about a year ago (?) that allowed you to updated a previous message

Arc: we briefly discussed the message replacing one, you'd have to fully replace the message and the new one would have to have its own offsets

Unless this was something different?

what if it only updated the markup, or added to it.

something like this would have to survive numerous tests, such as whiteboarding

yours is obviously simpler, and I agree that its a good thing.

frankly i think this whole issue is a bikeshed that's only going to rage on until everyone's exhausted and the issue is dropped again

im sitting next to our UX guy and he's hard eyerolling at the whole issue

I think that's just a waste of time and we have other more interesting things to tackle

i think tackling xhtml-IM is important. but yes there's so many other issues.

the construction workers still need their bikeshed.

Yah, really this whole thing only exists because for some reason people think we can't deprecate XHTML-IM until there's something else vaguely similar

i like that inline images are no longer included in either proposal.

SamWhited, I think deprecating XHTML-IM is not the answer in the first place and that's what started this whole holy war.

pep.: there's strong feelings on either side of it. but the core point remains, until someone has solved the core issue with xhtml-im it is *not* worth user security for the rarely used features it offers

xhtml-im has had a decade of real world use, and the security issues are yet to be solved. most don't even try. that's the issue.

Indeed :(

Well if they don't even try they're not going to try much more with other implementations. But I've already been told to shut up for whatever reasons

pep. that's why we need specs that just don't have the same fundamental flaws, so that people who just don't try (or who do try but make a mistake) are less likely to shoot themselves in the foot

or in this case, shoot all their users in the foot.

Then I guess XHTML-IM is not the only XEP you should tackle, I hope you have time for the rest of every single XEP that tries to display something to the user

we're getting slaughtered on the UX front.

By the way. Conversations now supports code blocks as well. The next stable release will probably be tomorrow or the day after. And then we will get some real world feedback on the styling thing

And more cluttury interface :( I already hate the blockquote thing

When I don't explicitely ask for it

pep.: i agree there are many. and I think there's a large interest in addressing them, xhtml-im is just a low hanging fruit with massive security issues

I think jonasw's proposal answers the security concerns fairly well though

It may still be improved, but that's a step forward

pep. yes, both address the concerns with XHTML-IM as far as I can see.

I don't dislike his proposal. I see merits in it. We're not going to resolve it today, there'll be some time to consider and discuss.

daniel: nice, that was quick. I still haven't built the branch so maybe I'll just wait now

Arc, as you said earlier, I only see this resolve itself when everybody gets tired of it and one of the sides gives way

(or maybe it wasn't exactly what you meant)

there's a number of young XEPs now that deal with ranges demarking substrings within message bodies

pep.: not exactly. you and I? we're not client authors.

daniel is. he and others have the final say, regardless of whatever is published in the XEP. ideally the XEP that gets promoted reflects the consensus that evolves out of this discussion.

I liked goffi's point btw, against the "You can have the plaintext version differ from the rich content if you include both".

Arc, well he gets the final say for his client

And his client only. That will probably influence other clients though, or the war might continue to rage, I don't know

pep.: do you author a client?

In the process of

oh neat, what platform is it for

everything we're working on is web-based, tho i have next to no involvement with the frontend.

We're still a bit far to say that, I guess it'll run on whatever the authors use primarily. The code is in Rust, I'm not sure about the frontends yet, we're sorting the xmpp stuff out first

great to hear about someone using Rust, its a very cool language

It is indeed :)

nice; my little toy client is also in Rust, but I've never gotten it far enough along to want to publish it.

its a shame they haven't trimmed it down for embedded use tho, I'm still waiting on that

Arc, there are quite a few use cases for it in embedded already from what I hear, but I have no experience in it

Arc: you can actually build without the standard library which is where all the stuff that relies on the OS lives and you end up with a minimal core library which works great on embedded hardware

SamWhited: oh really? do you have a link to that?

yah, just a moment

You heard about TockOS? Also as SamWhited says, have a look at https://github.com/japaric/xargo

I'm not sure if there's a section in the rust book yet, but the stripped down version of the standard library with no OS stuff is called "core"

There's a guy that's quite keen on doing embedded stuff in the Cambridge Meetup

Arc: https://doc.rust-lang.org/book/first-edition/using-rust-without-the-standard-library.html might help

https://github.com/thejpster/stellaris-launchpad

Cambridge UK or Boston?

UK

SamWhited: that's excellent, i'll run some tests later

oh, and can you bundle assembly with rust?

either in-line or linked in different source files

sorry im lazywebbing

You can, search for the "rust nomicon", I think that has a chapter

https://doc.rust-lang.org/1.12.0/book/inline-assembly.html

very nice.

Obviously all safety guarantees go out the window at that point :)

Rust is a cool language, I wish I had more reason to use it day to day

I wish I had more opportunities, reasons I have plenty of :P

SamWhited: things like bitpacking and hashing routines are things i regularly optimize for

well. "regularly" is an overstatement. but still.

SamWhited: move to Portland, lets give you reasons to use it daily :-)

Arc: I can't, I watched the TV show and now I can't go to Portland without laughing uncontrollably

Also, I'm reasonably sure that it's a fictional place and that no one actually lives in Oregon.

Portlandia is unfortunately closer to the truth than any of us are comfortable with, but it is a real place.

If that show is any indication it actually reminds me a lot of Austin

I've heard that as a comparison, but Portland is certainly weirder. For example, Austin doesn't have nude beaches that are a central social spot to just hang out

Is anyone else getting the permanent notification that Flow isnt typing anymore in conversations?

or this (NSFW) https://pdxwnbr.org in which naked people basically take over the city, walk into a random convenience store or cafe and you're likely behind someone in their birthday suit

Arc: hah, I've heard of that, sounds… interesting.

CTRL-H held an after-party this year where naked people took over the hacker space

In a hacker space full of laiths and laser cutters a naked after party just seems dangerous …

the main room is a big coworking space and lounge. that's where I am now. all we have here is an occulus rift and a few 3d printers, totally fine for parties

ctrl-h is 5 buildings and a shared backyard.

http://www.sheut.net/CTRLH-2017-11-07.jpg