Skip to content

Latest commit

 

History

History
979 lines (496 loc) · 32.1 KB

09-30-minutes.md

File metadata and controls

979 lines (496 loc) · 32.1 KB

W3C Technical Architecture Group face-to-face

- DRAFT -

September 30


Attendees

Present: Daniel Appelquist, Tim Berners-Lee, Domenic Denicola, David Herman (afternoon), Yehuda Katz, Sergey Konstantinov, Yves Lafon (remote), Peter Linss, Mark Nottingham, Alex Russell (afternoon), Jeni Tennison

Guests: Dominique Hazaël-Massieux (for discussion on Permissions)

Chairs: Daniel Appelquist, Peter Linss

Scribes: Mark Nottingham, Domenic Denicola, Sergey Konstantinov

Contents

Agenda


### Topic: EME

Domenic: Spec editor came to us at the last meeting to help us understand the current state of EME and answer our questions.

Domenic: I worked out a proposal to make EME more "webby" -- more conformant to shared values of privacy, platform independence

Domenic: I submitted a revised version of our opinion, stating concerns rather than solutions

Domenic: Sergey reviewed, accepted pull request, then sent a followup which we need to discuss

Domenic: Also, we got a reply from Mark Watson (Netflix) on list, which is good. His feedback was mostly about phrasing, softening things, etc. which I think we can address and is worth doing.

Domenic: It'd be great to issue this as a finding that can be referenced as a TAG opinion when closing bugs on the spec.

mnot [ group looks at document ]

twirl https://github.com/w3ctag/eme/blob/master/EME%20Opinion.md

Domenic: Sergey has an alternative design that he'll take to the group as potential future work.

Domenic: EME has momentum; it's shipping in most browsers, vendor-prefixed; we're just trying to limit the damage.

Domenic: We may still get EMEv2 or DRMv2 e.g., based upon public key cryptography, but that's a multi-year effort.

TimBL: I'd encourage Sergey to do that.

twirl Further reading: w3ctag/eme#8

twirl And http://lists.w3.org/Archives/Public/www-tag/2014Sep/0039.html

mnot [ group reviews pull request ]

Domenic: Sergey proposed sections on fraud prevention, accessability, fair use, anti-circumvention

Domenic: My opinion is that these are outside the scope of criminal activity

TimBL: I disagree

Peter: The whole point of DRM is to prevent criminal behaviour

TimBL: You don't have to say "fraud"; it could be "attack"

Domenic: That could be OK; "attack vector" is better

Sergey: I agree with Domenic; this shouldn't be limited to criminal behaviour. I'm 100% sure that some vendors will give you a license with additional restrictions. So let's make it clear that the user has to be able to read the license.

JeniT JeniT has joined #tagmem

TimBL: ... or you could leave affordances for regulators so that they can make appropriate limitations.

Domenic: We should be concerned about the user, and phrase it that way.

Domenic: The main point is that the user should be able to inspect the license, as this helps prevent attacks.

Domenic: "Fraud" is less concrete than "attacks"

ACTION Domenic and Sergey to re-phrase "fraud" as "attacks"

trackbot Created ACTION-876 - And sergey to re-phrase "fraud" as "attacks" [on Domenic Denicola - due 2014-10-07].

mnot Next - Accessibility

Domenic: I don't think this works. It's saying that in a hypothetical future when they can translate video, we should punch a hole in EME for that technology.

Domenic: I would rather approach this as saying that text tracks should not be encrypted. I have feedback that this is feasible and not highly controversial.

Domenic: If we want videos to be accessible, we'd need a huge societal / legal effort.

Peter: There are high contrast modes, e.g., those applicable to images.

Peter: The video stream shouldn't be lopped out of that, e.g., if the browser is in high-contrast mode.

Peter: A solution could be that the requirements can be pushed down.

Domenic: If we soften it to "The normal accessibility modes of videos should be available."

Sergey: [ draws diagram CDN - CDM - Hardware with pipe between CDM and Hardware to "System Service" ]

Domenic: System services could give instructions to CDMs to help accessibility

Peter: It could be a filter that has to be applied

Domenic: these are specific technical solutions; there could be others

Sergey: So we just add a note

Domenic: yes.

ACTION Domenic to add a note generally encouraging accessibility in EME opinion

trackbot Created ACTION-877 - Add a note generally encouraging accessibility in eme opinion [on Domenic Denicola - due 2014-10-07].

mnot Next - fair use

Sergey: Amazon displays books in the browser, and you can't copy it. The HTML was a complete mess of iframes, etc.

Sergey: In a short time, everything will be like that.

Domenic: I worked for bn.com and we did ebooks. E.g., you were granted 45 pages of copying every thirty days.

Torgo: Because it's important in an academic context

Domenic: yes

Torgo: The thing about fair use is that it's specific to jurisdiction, and varies quite a bit.

Torgo: IANAL

Torgo: AIUI the EU doesn't have a concept of Fair Use

TimBL: Yes they do

JeniT: It was never there as far as I'm aware, at least called "fair use"

mnot Torgo radiates joy

Torgo: There's work going on on a EU copyright overhaul

Torgo: My point was that it's not universal

Torgo: DRM stops users from doing things that are legal any way; e.g., excerpt of a video, which is legal under many copyright laws

Sergey: video companies think that all copying of videos is illegal

Domenic: Do we want to soften this to something like "We know that DRM and Fair Use have been in conflict, and would encorage discussion"

mnot [ derision erupts ]

Torgo: My thinking is that it'd be useful to say something like "DRM often foils people from doing something that's legal for them to do, according to their copyright law. This issue is not addressed in this document."

Domenic: That's only useful from the perspective that Tim brought up -- that future lawmakers might find it useful.

JeniT: Does that even make sense as a position though? Lots of tools don't let us do things that are legal for us to do.

Torgo: They restrict you from doing something that you could otherwise do on the Web, under the pretense that you can't do this because it's illegal.

TimBL: We talk a lot about UI quality and principles. In a way, to say we're introducing something that restricts the UI is a UI issue.

TimBL: What does the TAG think about web sites that add their own link and attribution when you copy?

Domenic: I think it's horrible

Peter: We can't technically restrict people from being annoying; we'd shut down the entire web.

JeniT: I think the usability aspect is important

Domenic: How can we say that constructively?

Peter: It's breaking the whole "view source" model on the Web

Domenic: That's a can of worms

Peter: There's a lack of developer tools; EME could prevent them from coming about

Domenic: That argument means "no CDM"

Peter: Or it means that the CDM exposes the content in a trusted way

Domenic: The point of DRM is that there is no trust.

Domenic: That might be a good motivator for DRMv2

Domenic: Kill it?

mnot [ group seems to agree ]

mnot Next - Anti-Cirumvention law

Domenic: This is an interesting one; not sure what to do with it.

mnot: what vulnerabilities would they be looking for?

Domenic: Like anything that ties into the OS

Peter: it has low-level access to the hardware

mnot: This seems like a purely legal problem

Domenic: agreed, but it's a good problem to bring up

TimBL: why is this a legal problem?

Domenic mnot: the problem is stated that there's legal exposure for people who try to do penetration testing. that's in the legal realm

Domenic timbl: so it's a problem that should be solved by changing the law, not changing the technology?

Domenic mnot: i don't see how else

Domenic timbl: well, if the whole setup was different (e.g. open source DRM), it would no longer be a legal problem

Domenic mnot: but that's a pretty radical thing to put into the document

Domenic timbl: agreed, not at this point, although a radical re-engineering is still a good idea over a longer timescale

Domenic timbl: but we should raise the concern

Domenic mnot: indeed, if only to give affordances to lawymakers later

Domenic: My intuition is that you couldn't design a spec to make it not possible to sue (sergey's suggestion)

Domenic: However, we could have something like an equivalent to the patent policy; "I won't go after pen testers"

Domenic: We'd need to talk to the AB about that

TimBL: AB might not be appropriate. It's worth pointing out that there are other possible protocols; it's half legal and half engineering

Domenic: I only suggest the AB because they have experience with the patent policy.

Domenic: I do like this feedback and suggest we do something with it; not sure where it belongs.

Torgo: I like the idea of a covenant.

mnot: +1

TimBL: Using SMTP as a model; suppose that the mail RFCs said that conforming use of the To: header is that it's used to give a functional address for the social entity that sends the e-mail. A conforming user is one that does these things...

TimBL: You then leave it open for regulators to use that.

Domenic: Making spec into certifications is a big thing

TimBL: That's different

Domenic: I think we should remove this from the document, but draft a communication to the AB asking for collaboration

mnot: Will it do harm in this document?

Torgo: Then we can point to this when we talk to the AB

Domenic: I think it needs to be separated. A lot of this is about concrete restrictions on the API; this is a separate concern.

Peter: This fits under the category of security concerns.

Domenic: good point

Domenic: We'd want to modify it into "We'll be investigating this..."

Peter: Say that this is a hole that we don't want to leave open

mnot: As long as we make it clear that we don't expect them to solve this problem.

Domenic: I appreciate the notion that this could be built on top of web crypto, but don't think it's directly relevant

mnot group resolves to rephrase, make it clear that this is an action for the TAG, not EME; make it clear that this is a security risk, put under security section ]

mnot Looking at Mark Watson's feedback

Domenic: There are four points.

Domenic: First - phrasing of "The ability of the CDM to potentially run arbitrary code is a hole in the web platform's security model."

Domenic: He has a point; e.g., Flash

Domenic: OTOH it's aritrary from the standpoint of the user, and opaque to the user

Domenic: We can either change from "arbitrary" to "unspecified" or add a sentence giving context

Sergey: With Flash, SWF are a set of instructions. With EME, we don't know that video is just data, not code.

Domenic: So restrict what data goes in and out?

Sergey: Yes, so only data goes in and out

Domenic: That seems to go against years of computer science

Domenic: Oh, you're saying that frames shouldn't smuggle code?

Sergey: Yes

Domenic: There is no restriction that it is video presently?

Sergey: They're talking about using standardised form of encrypted video. But the license response could contain instructions for the CDM.

Domenic: I think this is worth bringing up in our reply to Mark; it may not change what we put in the document though. I haven't talked with many people about the license response.

Domenic: Second concern - "As part of interoperability, EME should not provide APIs that are designed to allow restriction of content to one platform and/or key system."

Domenic: We should do a more thorough investigation by what is meant by this.

Peter: Isn't it the opposite, that EME is failing to provide universal APIs?

Domenic: I think we covered that pretty well.

Domenic: More investigation needed.

Domenic: Third concern - "While certain key systems may only be supported on certain platforms, and certain content may only be available with certain key systems..."

Domenic: I'm not sure what he's concerned about here.

mnot: he seems to be just noting something.

Domenic: Fourth concern - "We understand that designing a DRM system for the web brings along robustness requirements that are unlike those of most Web APIs..."

Domenic: My take is that this is not very relevant.

TimBL: He's saying "Don't focus on robustness."

TimBL: Maybe he's saying that people should be allowed to think about systems that don't have robustness

Domenic: I want to push back because I don't see this expressed in the EME document

mnot: We're not writing a requirements document; it doesn't have to be complete

Domenic: Maybe he's concerned that we're mischaracterising it

mnot [ side conversation between Domenic and Sergey ]

Domenic: Tim's point is that we shouldn't say "necessary for robustness" as if that were a feature for all DRM

Domenic: ... but EME has chosen to address DRM in this way

Peter: You could address this by removing the word "robustness" from that sentence

Domenic: OK

Domenic: I like this, because it addresses Tim's and mark's comments

Domenic: Or say "Robustness or IPR"

Torgo: I like Peter's suggestion more

Torgo: Keep silent on the motivation

Domenic: That's it for EME

timbl http://tag.w3ctag.org

plinss https://pad.w3ctag.org/

timbl

timbl The web address you've requested doesn't exist

timbl The web address you have requested is incorrect or does not exist. You could try amending the spelling or check the web address.

timbl

timbl Pppht

mnot [ group agreed to send this as spec feedback ]

timbl https://pad.w3ctag.org/p/eme

Domenic { "&" : "&"} ===> <json:string name="&">&</json:string>

### Topic: Captive Portals

mnot: we've talked about captive portals a lot in the IETF and elsewhere

mnot: if we wanted to try to do something, we could probably get the right people from places like Cisco, Microsoft, Apple, the browsers, ...

mnot: where we're at now, is, "what can be realistically done?"

mnot: we've had preliminary discussions mapping out the problem space, figuring out back-compat constraints

mnot: big question is what is the right layer to interpose these things

mnot: we have a status code in HTTP for this, but you can't intercept that when you go over HTTPS

mnot: anecdotally I've been seeing fewer captive portals, just WiFi passwords

Domenic (the room disagrees)

mnot: a common use case is just showing some text before a user connects

Yves well, ssl should show certificate issues if connected to a captive portal

Domenic (discussion of various problems with bad cert errors upon connecting to a network and how this is only getting worse)

timbl: i could see OS designers lying, saying "there is no network," until they can detect actual internet

mnot: yes, but there is a war between the networks and the OSs, some captive portals want to overcome captive portal detection

dka: this is why apple has 100 different domains to detect this...

mnot: it's only getting worse because Apple is doing MAC address randomization, causing captive portals to lose their ability to remember you

mnot: low hanging fruit would just be documenting what OSs do right now

timbl: what would it look like to design this from scratch?

mnot: we need a transition stategy

timbl: no, I'm more interested in benevolent places like MIT

mnot: my gut is if you wanted to do this from scratch it would be part of 802.11

plinss: what about DHCP?

mnot: we're trying to get away from that; inherently insecure

plinss s/plinss:/timbl:/

timbl: "here's your IP address; it's not going to work until you visit this web address"

Yves DHCP is insecure, but still you configure your interface (and dns settings) using dhcp replies

mnot: but because of how DHCP works, anyone can announce that to you, so it's a great way to phish someone in a coffeeshop

timbl: are we ratholing?

mnot: the problem is captive portals are full of ratholes like this because it's full-stack

timbl: i'm interested in solving this for enterprises and universities...

mnot: well universities have alreayd solved this with e.g. eduroam...

plinss: (but it's still a crappy implementation)

dka http://www.btplc.com/news/articles/showarticle.cfm?articleid=%7B2b5f7e38-2551-47d6-9d01-21f44cb25d26%7D

plinss: opportunity to invite some of the other parties around TPAC?

mnot: I was thinking more a full-day workshop. And TPAC is only a couple weeks away

mnot https://github.com/httpwg/http-extensions/wiki/CaptivePortals

timbl timbl has joined #tagmem

plinss: I'd like to see an end result of "ideally, here's a spec for a good captive portal"

Domenic (discussion of a "brand" for well-designed captive portals and related stuff)

mnot: my gut feeling is that this is a long-term thing

mnot: perhaps an event of early to mid next year

plinss: perhaps adjacent to EWS in April

mnot: concrete next steps---work on the wiki, talk to people to set up a meeting, maybe come up with a plan for a plan

Zakim Zakim has left #tagmem

timbl Hitler reacts to the HTML5 URL normative reference controversy - See more at: http://meemsy.com/v/22036#sthash.I7RqFP9V.dpuf

timbl The last bit was snuck in by the site — how?

twirl ScribeNick: twirl

dka Starting Up Again

### Topic: Tooling / Github / Etc…

twirl mnot:

twirl mnot:

twirl mnot:

twirl mnot: Markdown rules

twirl mnot: The hardest decision is doing something with our home page

twirl dka: I think it's a high priority task

twirl dka: EME is a good example, it's actually spec review but it's not in spec reviews repo

twirl mnot: Chairs should control repo creation

twirl [ disagreement ]

twirl [ mnot is very persuasive ]

twirl [ discussion on whether github wiki is better than TAG Wiki or not ]

twirl [ discussion on using wiki vs repos ]

twirl mnot: Repos could be used offline

<bkardell_> bkardell_ has joined #tagmem

twirl mnot: And github wiki too :)

twirl mnot: github wiki is a separate repo

twirl [ mnot explains how HTTP2 repos are organized ]

twirl [ discussion on organizing indexes ]

twirl dka: proposes (a) move to tag.w3.org -> w3tag.github.io, (b) move meeting organization to github, (c) move minutes to github

twirl (d) move all current work to main page, (e) move EME to Spec Reviews repo, (f) clean up repos, (g) backup everything, (h) elaborate policy on repo creation

twirl mnot: are we ok with breaking URLs?

twirl timbl: I think its important to preserve URLs

twirl s/its/it's

twirl [ dka explains our current policy ]

twirl plinss: let's proxy all URLs into W3C space

twirl [ timbl explains how to convert cvs subtree ]

twirl [ plinss explains how to synchronize github and w3c space ]

twirl plinss: Minutes URLs MUST be preserved

twirl dka: Having our documents in github, having clear repos are good ideas

twirl dka: But I'm very concerned about preserving URLs history

twirl [ general discussion ]

twirl dka: We could put on our home page everything which changes rarely, and have a separate gh-page for current work

twirl mnot: I could try to convert our old stuff

twirl [ discussion on CVS -> git conversion ]

twirl mnot: what about gh issue tracker?

twirl mnot: two trackers are confusing

twirl [ historical joke from dka ]

<bkardell_> ...But GH is very Dev friendly

twirl dka: the problem is that each repo has its own tracker, so you can't see all w3c tag issues

twirl mnot: let's use Arch tracker for long-living issues

twirl [ discussion on confusing trackers ]

JeniT all issues: https://github.com/organizations/w3ctag/dashboard/issues/repos

twirl dka: let's setup a repository for "TAG" issues

twirl dka: and close w3 issues

twirl mnot: what to do with spec reviews wiki page? let's put in README in specreviews repo

twirl [ dka tells the story about agendas in cvs ]

twirl dka: let's move agendas to github

twirl mnot: what's the metrics of success?

twirl [ memes, videos, number of documents ... ]

### Topic: Spec Reviews

twirl Domenic: we were asked to review WebApps specs

twirl Domenic: WebCrypto is done

twirl twirl: Web Animations review is done

twirl mnot: HTTP2 reviewed

twirl [ reviewing w3c issue tracker ]

twirl mnot: Get off My Lawn: closed

twirl http://lists.w3.org/Archives/Public/www-tag/2014Jun/0004.html

twirl http://lists.w3.org/Archives/Public/www-tag/2014Jun/0002.html

twirl mnot: CSS Regions?

twirl [ plinss explains elements flows ]

twirl dherman: Is there any progress with box trees?

twirl plinss: yes, we are moving

<bkardell_> Was a task force set up?

Yves good also to put the discussed scope in the issue description

twirl [ looking at: CSP2, Mixed Content, Referrer Policy, ... ]

twirl Domenic: I've reviewed Referrer spec, found some issues

twirl [ mnot and Domenic take this issue ]

twirl [ Domenic explains SRI problem ]

twirl mnot: there severals spec related to URL metadata

twirl s/spec/specs

twirl s/there/there are

twirl w3ctag/design-reviews#36 (comment)

twirl Domenic: best way to do spec reviews is talking to author directly

dherman "on their turf" is a powerful point

dherman "TAG is here as a service" <== good attitude

twirl [ Screen Orientation is done ]

dherman "come over to my house and let me tell you how you're wrong" <== bad attitude

dherman TAGaaS

twirl [ multipart/form-data ]

twirl mnot: it's already Last Call

JeniT http://www.ietf.org/rfc/rfc2388.txt

JeniT (presumably the previous version)

twirl Domenic: looks cool, there are tests

twirl [ work suddenly stopped ]

twirl w3ctag/design-reviews#34 (comment)

https://github.com/masinter/multipart-form-data/issues/17#issuecomment-48077041

twirl [ reviewing multipart-form-data issues ]

twirl [ off-topic: mnot complains about implementing HTML5 ]

twirl [ agreed to propose something by TPAC ]

twirl mnot: Wake Lock

twirl Domenic: I'm already involved

twirl I'd prefer to have a simple API

twirl s/I'd prefer/Domenic: I'd prefer

twirl http://w3c.github.io/wake-lock/

twirl [ reviewing proposed use-cases ]

mnot: Full Screen: closed

twirl mnot: Manifest?

twirl Domenic: It's for installable webapps; related to packaging

twirl mnot: WebRTC Identity Provider Selection

twirl Domenic: bizzare thing, something like invisible iframe

twirl w3ctag/design-reviews#28 (comment)

slightlyoff OMW

slightlyoff hey all

slightlyoff terribly sorry

slightlyoff meant to join 2 hours ago and failed = \

dom dom has joined #tagmem

slightlyoff worse in many, many ways = )

### Topic: Permissions

slightlyoff terribly sorry for how long that took

dka https://gist.github.com/slightlyoff/43cd8c2f64a0719358fe

Alex there are several key aspects

... you can request a permission at page load

... you may build UI using API calls results

... all these things are async

... and user can always say no

... we can imagine lighter UI than prompts if user trusts that app

... we have lots of experience with mobile apps sync permission dialogs, and its disappointing

... there is a trade of remembering webapp permissions vs. leaking this information

twirl: how to revoke permissions?

Alex: I separated UI from API in a document

twirl: but how will webapp know its permission revoked?

Alex: ask each time

Domenic: but how to restyle UI?

Alex: probably perission change event

Dom: I like this proposal

... we discussed those questions, I think it's a good starting point

[ discussion on requests ]

Alex: the question is whether it reasonable to create separate API requests to each action

mnot: there is no easy answer, in mobile APIs too

Alex: you should have a freedom to change your mind

Alex: we could do that in extensible way, with low-level API

dom: It would be great if TAG provides more guidance on securing the Web

slightlyoff I am still working on the origin post, it's much harder

slightlyoff so I have an idea of where we're going here

slightlyoff I can outline it briefly

dom: We need a clear picture of future web security

Alex: there are good success stories from OSes, we should work on browser UI

... we should think about "installation" for webapps

slightlyoff I don't think WG's are the right place to be litigating these issues

dom: we need even broader picture

slightlyoff nearly all the important security UI is owned by the UA

Domenic permissions, origins, crypto, UI

slightlyoff and that's increasing over time

Domenic (i really like this idea that we should lay out the landscape of how all those things interact)

Alex: we now have some legacy, files, we can't ask, for example, a single contact information

dka ok thanks Dom!

timbl: I need to understand why app wants to have some particular access, I have to trust it completely

Alex: we may think for example about manifest, so user could understand what application is

twirl ... I don't believe we have answers; for example, should permissions be granted to app working in iframe, etc

twirl s/answers/all answers

twirl [ timbl on peculiar requests bouncing in Chrome ]

dka dka has joined #tagmem

twirl [ discussing the problem of sending credentials with CORS ]

Alex: Credentials are the most valuable permission, more sensitive than raw socket access

timbl: imagine that I'm trying to write really good calendar, I need to be able to get information from all other sources

twirl ... at present moment ony Facebook could be such super application

twirl ... that's reputation question

Alex: my experience tells that user doesn't understand what really happens

twirl [ discussion on service workers which help to solve these problems ]

dka: Dom asked could we document security model beyond permissions

timbl timbl has joined #tagmem

twirl ... and that's an action for the TAG in my view

Alex: I've been working on origins post

twirl ... we need to evaluate question how much more could you trust smth, there is no answer right now

twirl ... Service Worker is a thing which controls origins since it sees all requests

mnot: we have same problems with proxies

twirl [ Service Worker Union, we need it ]

timbl http://www.seiu.org

timbl clientside proxy

slightlyoff timbl: we avoided that because it doesn't describe the other background stuff we can do with it, e.g. responding to Push messages

slightlyoff timbl: also, I'm going to look into the XHR/fetch credentials thing and see what conditions we can relax it under

### Topic: Spec Reviews (continued)

slightlyoff thanks for bringing it up

mnot: Shadow DOM

twirl [ awwwww ]

mnot: Closed!

mnot: Navigation Error Logging

mnot: They added promises

mnot: I'll check implementation status

mnot: Beacon

Alex: API is ok, but there could be some implementation problems

mnot: NFC API

slightlyoff this API doesn't look half bad!

twirl so many abbreviations

slightlyoff questions for NFC: what about "background" NFC? integration with SW's?

mnot: CSS Font Loading

Domenic: nearly done

mnot: Web MIDID

twirl s/MIDID/MIDI

mnot: Closed

mnot: WebRTC

slightlyoff I think we should perhps try to get the ORTC folks to present?

slightlyoff maybe have them talk to us?

dka: let's organize a call

slightlyoff thanks

slightlyoff I would also like to let both "sides" know that we're paying attention

slightlyoff and would like to see resolution + interop

mnot: Web Components

Domenic: done!

mnot: Push API

Alex: already done, feedback accepted

mnot: Web Audio

Domenic: done

mnot: Compositing and Blending

mnot: Closed

slightlyoff the only thing about Push that's nasty is the verbosity in the API: https://w3c.github.io/push-api/#example

slightlyoff but otherwise I think it's in good shape

twirl verbosity is better than letter-saving

Domenic: next time we should ask authors about preferred form of communication

slightlyoff the call is SUPER static-y now

slightlyoff gonna re-dial

slightlyoff thanks

### Topic: Privacy

mnot: I talked with Chrome and HTTP teams, they were frustrated with Privacy mode [not] understood by users

mnot: people don't understand what Privacy Mode guarantees and what does not

mnot: if we standardize definitions of privacy mode it will help

mnot: users and websites will learn what it means

mnot: it's probably a proper moment to act

slightlyoff link?

mnot: I tried to define what browser privacy mode is

mnot: hiding clues from other users of the computer

slightlyoff I'd LOVE for sites to get unique storage in this mode

mnot: hiding from external observers

mnot: hide identity from websites

slightlyoff the measurement APIs are an area we can do better in a privacy mode

slightlyoff also font access

dka: fingerprinting is important too; I can be okay with tracking in regular mode, but I'd like to be not tracked in privacy mode

plinss: but you can, in fact, identify yourself as a person-which-is-untraceable

mnot: browsers should be more agressive in failing situations

mnot: and last one: get user informed

mnot: in privacy mode browser either should either obfuscate traffic or display an alert

Domenic: browsers should compete how many privacy features they implement

Alex: I think in privacy mode access to storages should be restricted

dherman: how could this affect the platform?

slightlyoff specifically I was saying that you might be able to partition storage

mnot: spec editors will be aware of several browser modes

dherman: idea itself seems fruitful, but AFAIK users use Privacy mode for hiding browser history, not to be not tracked

dherman: we should not make decisions based on our view, not real people needs

dherman: it's a product decision, not platform

timbl: that's just a switcher saying "I will be embarassed if anyone knows"

timbl s/knows/knows — it is up to the tech community t try to anctipate the attack vectors, not the user/

dka: there is not enough knowledge [in the public sphere] about when it's a good idea to use privacy modes [i.e. not only when you are trying to "hide" something]

twirl s/there are too less knowledge about/people know very little

JeniT: does this affect server-side?

mnot: in general not; maybe in a form of profiles and cookie politics

dka adjourned

slightlyoff heading off

slightlyoff later all!

dka thx!