- From: Ben Laurie <benl@google.com>
- Date: Mon, 1 Oct 2012 17:02:04 +0100
- To: Henry Story <henry.story@bblfish.net>
- Cc: "Jonas Hogberg K.O" <jonas.k.o.hogberg@ericsson.com>, Carvalho Melvin <melvincarvalho@gmail.com>, "public-philoweb@w3.org" <public-philoweb@w3.org>, "public-webid@w3.org" <public-webid@w3.org>, Oshani Seneviratne <oshani@mit.edu>
On 1 October 2012 15:36, Henry Story <henry.story@bblfish.net> wrote: > > On 1 Oct 2012, at 15:54, Ben Laurie <benl@google.com> wrote: > >> On 1 October 2012 14:51, Henry Story <henry.story@bblfish.net> wrote: >>> >>> On 1 Oct 2012, at 15:46, Ben Laurie <benl@google.com> wrote: >>> >>>> On 1 October 2012 14:41, Henry Story <henry.story@bblfish.net> wrote: >>>>> >>>>> On 1 Oct 2012, at 15:36, Ben Laurie <benl@google.com> wrote: >>>>> >>>>>> On 1 October 2012 14:07, Henry Story <henry.story@bblfish.net> wrote: >>>>>>> >>>>>>> On 1 Oct 2012, at 14:35, Ben Laurie <benl@google.com> wrote: >>>>>>> >>>>>>>> On 1 October 2012 13:20, Henry Story <henry.story@bblfish.net> wrote: >>>>>>>>> >>>>>>>>> On 1 Oct 2012, at 13:43, Ben Laurie <benl@google.com> wrote: >>>>>>>>> >>>>>>>>>> On 30 September 2012 20:22, Henry Story <henry.story@bblfish.net> wrote: >>>>>>>>>>> >>>>>>>>>>> On 30 Sep 2012, at 20:46, Ben Laurie <benl@google.com> wrote: >>>>>>>>>>> >>>>>>>>>>>> On 30 September 2012 10:30, Henry Story <henry.story@bblfish.net> wrote: >>>>>>>>>>>>> >>>>>>>>>>>>> On 29 Sep 2012, at 19:50, Ben Laurie <benl@google.com> wrote: >>>>>>>>>>>>> >>>>>>>>>>>>>> On 28 September 2012 15:26, Jonas Hogberg K.O >>>>>>>>>>>>>> <jonas.k.o.hogberg@ericsson.com> wrote: >>>>>>>>>>>>>>> At >>>>>>>>>>>>>>> http://blogs.kuppingercole.com/kearns/2012/09/25/in-search-of-privacy/?goback=.gde_3480266_member_168314336, >>>>>>>>>>>>>>> Dave Kearns writes: >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> There is indeed a lot of confusion about the subject, but there are two key >>>>>>>>>>>>>>> phrases to remember when talking about privacy: >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> Privacy is not anonymity >>>>>>>>>>>>>>> Privacy is not secrecy >>>>>>>>>>>>>> >>>>>>>>>>>>>> Quoting those out of context is not particularly helpful. But for more >>>>>>>>>>>>>> on why anonymity is important for privacy... >>>>>>>>>>>>>> >>>>>>>>>>>>>> http://www.links.org/?p=123 >>>>>>>>>>>>>> http://www.links.org/?p=124 >>>>>>>>>>>>> >>>>>>>>>>>>> Looking at those two, can we agree that we agree that anonymity should be the default? >>>>>>>>>>>>> I believe as you do that when I go to a web site the default should be that I not be >>>>>>>>>>>>> identified, and not be tracked. I can choose later to be tracked or identified for >>>>>>>>>>>>> that site for a given amount of time or until I change my mind, but the default should >>>>>>>>>>>>> be anonymity. >>>>>>>>>>>>> >>>>>>>>>>>>> ( Within limits of logic of course. If I tell anonymous Y something P >>>>>>>>>>>>> which has consequence Q, and some other anonymous Z does something with Q that would have >>>>>>>>>>>>> been nearly impossible to know had they not known P, then I could conclude within >>>>>>>>>>>>> a certain probability that Y == Z ) >>>>>>>>>>>>> >>>>>>>>>>>>> The web provides this. Some browsers provide it better than others, but really >>>>>>>>>>>>> this is up to them. It is not perfect: ip addresses can be tracked and dns lookups >>>>>>>>>>>>> can be tracked. But the web is not reliant on those. It could be deployed just as well >>>>>>>>>>>>> on top of Tor. Had people had better memories, we could have had .onion urls plastered >>>>>>>>>>>>> on bus stops since the beginning. >>>>>>>>>>>>> >>>>>>>>>>>>> Anonymity is important for many reasons. Among which is that it helps create a trusted >>>>>>>>>>>>> public sphere. It increases my trust in the information I read if I know that the publisher >>>>>>>>>>>>> publishes that information that can be read by anonymous readers. Knowing that the publisher >>>>>>>>>>>>> cannot tell who is reading what he is publishing is a very strong guarantee that he >>>>>>>>>>>>> is not adapting his message to different groups. Oddly enough anonymity has an important role >>>>>>>>>>>>> therefore in public discussion. >>>>>>>>>>>>> >>>>>>>>>>>>> So do we agree here? I think we do. >>>>>>>>>>>> >>>>>>>>>>>> So far. >>>>>>>>>>> >>>>>>>>>>> ok. So let's see if we can agree further, from here :-) >>>>>>>>>>> >>>>>>>>>>> There are a number of identification options available. >>>>>>>>>>> Let me list some of them: >>>>>>>>>>> >>>>>>>>>>> - anonymous ( 0 identification ) >>>>>>>>>>> - cookies ( site bound ) >>>>>>>>>>> - TLS-Origin-Bound-Certificates ( unforgeable cookies ) >>>>>>>>>>> - Self-Signed certificates with an .onion WebID >>>>>>>>>>> ( I promised Appelbaum to work on that. This gives you an identity, but nobody knows >>>>>>>>>>> where you or your server are located ) >>>>>>>>>>> - Self-Signed certificates with a http(s) WebID >>>>>>>>>>> - CA Signed Certificates >>>>>>>>>>> - DNSSEC Signed Certificates >>>>>>>>>>> - ...? >>>>>>>>>>> >>>>>>>>>>> We agree that anonymous should be the default. >>>>>>>>>>> I think we can agree as a matter of simple fact that none of the browsers show >>>>>>>>>>> you which of those modes you are in when looking at a web page. You cannot >>>>>>>>>>> as a user therefore tell if you are anonymous or not. You cannot therefore tell >>>>>>>>>>> if the page you are looking at has been tweaked for you or if it would appear >>>>>>>>>>> differently to someone else in the same mode as you. You cannot tell if the >>>>>>>>>>> agent on the other side can tie you to a browsing history or not. >>>>>>>>>>> >>>>>>>>>>> Well let me put this in a more nuanced way: you can tell the above from the >>>>>>>>>>> side-effects - say if they should you your profile on a google+ page with edit mode >>>>>>>>>>> allowed - but that is up to the server to show you that. We both want it to be >>>>>>>>>>> up to the user. We don't want it to be up to the user in some complicated conf file >>>>>>>>>>> hidden away somewhere. We both want it to be in your face, transparent. I should >>>>>>>>>>> in an eyeblink be able to tell if I am anonymous or not, and I should be able >>>>>>>>>>> to switch from one mode to the next if and when I want to in a simple easy gesture. >>>>>>>>>>> >>>>>>>>>>> Just as in real life when we put on a mask we know that we are wearing the mask, >>>>>>>>>>> so on the web we want to know what mask we are wearing at all times. >>>>>>>>>>> >>>>>>>>>>> These are the improvements I have been fighting ( not alone ) to get browsers to >>>>>>>>>>> implement. Are we fighting on the same side here? >>>>>>>>>> >>>>>>>>>> I agree that it is desirable to know how your browser is identifying >>>>>>>>>> you and to be able to switch between users. So, I guess Chrome would >>>>>>>>>> claim that the facility to have multiple users provides this. Do you >>>>>>>>>> disagree? >>>>>>>>> >>>>>>>>> I looked up multiple Users and found this: >>>>>>>>> http://support.google.com/chrome/bin/answer.py?hl=en&answer=2364824 >>>>>>>>> I had not seen this before. >>>>>>>>> >>>>>>>>> So it seems to work for certificates. I created a new user Tester, and >>>>>>>>> noticed the following as that Tester: >>>>>>>>> >>>>>>>>> 0. It did not have any of my bookmarks ( I suppose that's useful, cause your >>>>>>>>> bookmarks could identify you ) >>>>>>>>> 1. When I went to Google+ it did not know I was >>>>>>>>> 2. Having signed in to https://my-profile.eu/ as the old user, I tried as the >>>>>>>>> new user Tester, and had to select a certificate again. Good. >>>>>>>>> >>>>>>>>> So that seems like one way to separate one's personalities. I'd still like to >>>>>>>>> have the url bar show me for each tab: >>>>>>>>> >>>>>>>>> [anonymous] when I am not logged in >>>>>>>>> [cookie] when I am tracked on that site >>>>>>>>> [henry story] for a local site identity >>>>>>>>> [bblfish@home] when I am using a certificate >>>>>>>>> >>>>>>>>> With the option of logging out from that site (ie checking x -> anonymous ). Because >>>>>>>>> currently I could forget that I had chosen a certificate on a site, and it >>>>>>>>> would continue sending it. Or I could mistakenly choose a certificate as one user, >>>>>>>>> and then decide that was the wrong user for that persona, and not be able to choose >>>>>>>>> the certificate again, without closing my browser completely. That would allow, on >>>>>>>>> browser startup, the browser to remember the last identity choice for a site. Without >>>>>>>>> logout capability that is not possible, because then it would be impossible to repair >>>>>>>>> an identity mistake without creating a new user. (And it makes testing tedious). >>>>>>>>> >>>>>>>>> Currently when I close my browser, on restart the servers ask me for my certificate again. >>>>>>>>> >>>>>>>>> So it looks like this is going generally in the right direction. It still does not provide >>>>>>>>> the transparency we are looking for at the UI level above. But thanks for pointing this out. >>>>>>>>> >>>>>>>>> So I think we agree that what is missing is the transparency at the UI level of which identity >>>>>>>>> one is using at each site. That is what I was hoping the following bug report would achieve. >>>>>>>>> >>>>>>>>> http://code.google.com/p/chromium/issues/detail?id=29784 >>>>>>>>> >>>>>>>>> So perhaps by putting this forward under the term transparency, that would help that bug report >>>>>>>>> progress, since otherwise they could thing that the issue had already been completely solved. >>>>>>>>> >>>>>>>>> So that's what I make of that. But have I missed something? Or do we agree there too? >>>>>>>> >>>>>>>> I don't think so >>>>>>>> . As I said, I think that Chrome would claim that the >>>>>>>> users facility provides everything you need - if you want to know >>>>>>>> which cert you're using, then have a user per cert. As for cookies and >>>>>>>> "local site identities", this would require information the browser >>>>>>>> does not currently have, so I think you would first have to explain >>>>>>>> how it is going to get that information. >>>>>>> >>>>>>> Well the browser knows when it sends a cookie. So showing a [cookie] >>>>>>> icon would be easy there. When you are in anonymous mode it does not >>>>>>> send a cookie. (perhaps a no-cookie/cert icon - would be more precise) >>>>>>> As for per site identity that is what the Mozilla folks were working >>>>>>> with Aza Raskin >>>>>>> >>>>>>> http://www.azarask.in/blog/post/identity-in-the-browser-firefox/ >>>>>>> >>>>>>> But until a standard is agree to there, one could already have >>>>>>> a [cookie] icon... >>>>>> >>>>>> Sure, but it would be pretty pointless: I just checked and every >>>>>> single tab I have open has some cookies associated. >>>>> >>>>> So perhaps then only show anonymous when no cookie is there. >>>>> >>>>>> >>>>>>>> For anonymous, Chrome already has an anonymous mode (though note that >>>>>>>> you don't really stay anonymous for long once you enter it, since it >>>>>>>> must still use cookies or the 'net stops working - also bookmarks are >>>>>>>> still available in anon mode). >>>>>>> >>>>>>> As above the browser knows when it sends cookies: and so it can show >>>>>>> the user that it is doing that. >>>>>>> >>>>>>>> >>>>>>>> I believe that Chrome experimented with per-tab personas and found >>>>>>>> that it was a terrible user experience, btw. >>>>>>> >>>>>>> It does not look that bad in Aza Raskin's proposal, and the Account >>>>>>> Manager work at Mozilla >>>>>>> >>>>>>> https://wiki.mozilla.org/Labs/Weave/Identity/Account_Manager >>>>>>> >>>>>>> My guess is that the project to create the multiple user work >>>>>>> at Chrome trumped the development of good identity transparency >>>>>>> solutions. That often happens in engineering: one good idea >>>>>>> hides another one for a while. >>>>>> >>>>>> Or, as I said, it turns out to not work very well. That happens even >>>>>> more often, and apparently has happened in this case. Saying it >>>>>> doesn't look that bad to you doesn't change it! >>>>> >>>>> Look if we are serious thinkers we first select our principles and >>>>> then we search for a solution. It may be that we have not found the >>>>> solution, yet. But since we have established an important principle of >>>>> transparency, we keep looking until we find the solution. I am >>>>> not dictating the solution. I am saying we agreed on a principle, >>>>> so it is now a question of solving it in good will. >>>>> >>>>>> >>>>>>> In any case there is a lack of transparency in the multiple user >>>>>>> set up that still needs to be rectified. How that is done I'll leave >>>>>>> to UI experts. But I'll recognise a good solution whatever form it >>>>>>> takes. >>>>>>> >>>>>>> Now here with WebID we are assuming such a solution will be found >>>>>>> by one of the browser vendors in good time, and then adopted by the >>>>>>> others. The current interface we can agree is not good enough for >>>>>>> sure, but the problems we are trying to solve are important enough >>>>>>> that we can work with the current limitations of browser. >>>>>> >>>>>> Who is the "we" that can agree it? And why is it not good enough? You >>>>>> have not explained that at all. >>>>> >>>>> I did explain it. But it must have gotten lost in some threads. >>>>> I'll start a new thread on that. >>>> >>>> Specifically, I am asking why the users facility that Chrome has is >>>> not good enough... >>> >>> Because I cannot tell: >>> >>> - when I am anonymous ( as opposed to being tracked without my knowing it) >>> - what identity I am using when on that site: and this is just as valid for >>> cookie identification as for certificate identification. I can have multiple >>> profile accounts associated with different cookies. I can have multiple >>> identifying certificates. I want _my_browser_ to tell me which one I am using, >>> and not have to rely on the server, which may have more or less good >>> implementations for this. >> >> The idea is you create a user per identity. Then the browser is >> telling you which you are using. > > But that does not solve the problem. Because you could go away from > your computer, someone log you in under a different name, and you not > know. If you let someone else control your computer, I think it is game over anyway. > Or you could have logged in accidentally under the wrong identity. Hmm ... how? If the identity is linked to the user, you can't use the wrong one - that's rather the point. > Or......... > The browser should tell you what it is sending, because the browser > is what you control. Not arguing that this would not be nice, but I find it hard to imagine a usable interface - for example, you can use Chrome's developer tools to look at cookies that get sent, and its a pretty complex mess... > As user you don't control the server. This is the > transparency requirement. > > >> >>> >>>> >>>>> >>>>>> >>>>>>> That leaves us with the importance of cross site identity. I think >>>>>>> I have a very powerful argument in favour of its importance. It is >>>>>>> important for a certain kind of privacy to be possible: that between >>>>>>> two people or groups of people wishing to exchange documents that >>>>>>> should only be visible to certain people and no others. This is the >>>>>>> case when someone wishes to discuss something with a doctor, or when >>>>>>> someone wishes to publish photos of people at a party without making >>>>>>> it fully public, and in many many other circumstances. It is important >>>>>>> for creating a distributed social network, which I will call the >>>>>>> Social Web. The Web and the internet have always been about distribution >>>>>>> and decentralisation of information. We want to do that using WebID in >>>>>>> a manner that increases privacy. I will be working on showing how >>>>>>> this can be done on the Web, and on the Web running over Tor. >>>>>>> >>>>>>> Henry >>>>>>> >>>>>>> Social Web Architect >>>>>>> http://bblfish.net/ >>>>>>> >>>>> >>>>> Social Web Architect >>>>> http://bblfish.net/ >>>>> >>> >>> Social Web Architect >>> http://bblfish.net/ >>> > > Social Web Architect > http://bblfish.net/ >
Received on Monday, 1 October 2012 16:02:37 UTC