- From: Ben Laurie <benl@google.com>
- Date: Mon, 1 Oct 2012 14:54:04 +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 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. > >> >>> >>>> >>>>> 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/ >
Received on Monday, 1 October 2012 13:54:36 UTC