- From: Ben Laurie <benl@google.com>
- Date: Mon, 8 Oct 2012 13:04:43 +0100
- To: Henry Story <henry.story@bblfish.net>
- Cc: Melvin Carvalho <melvincarvalho@gmail.com>, "public-webid@w3.org" <public-webid@w3.org>
On 8 October 2012 13:01, Henry Story <henry.story@bblfish.net> wrote: > > On 8 Oct 2012, at 13:33, Ben Laurie <benl@google.com> wrote: > >> On 8 October 2012 10:53, Henry Story <henry.story@bblfish.net> wrote: >>> >>> On 8 Oct 2012, at 11:36, Ben Laurie <benl@google.com> wrote: >>> >>>> On 6 October 2012 08:48, Melvin Carvalho <melvincarvalho@gmail.com> wrote: >>>>> WebID is actually 2 specs. >>>>> >>>>> 1. The first part is authentication via your public key which is a IFP of >>>>> your identity. In certain circumstances (ie caching, just like >>>>> ~/.ssh/authorized_keys ) you can be done here and it operates like SSH. >>>>> >>>>> (1) I think solves the unlinkability problem >>>> >>>> How? Clearly the public key makes all authentications that use it linkable. >>> >>> +1 yes. >>> >>> It' is only unlinkable in the bizarre sense (which may in fact be nonsense) >>> in which for Harry Halpin BrowserId has some unlinkable properties that >>> WebID lacks. >>> >>> In any case the linkability issue is one which requires one to decide who >>> the attacker is according to the definition [1] >>> >>> • If the attacker is the site you are logging into, and you want to communicate >>> with that site unlinkably - as a wikileaks leaker would want to do to cover his >>> tracks - then using WebID, BrowserId or other such systems is really not the >>> right technology. That is pretty self evident. >>> >>> • On the other hand if you want to avoid a centralised network, be able to create >>> long term relationships across organisations - and the site you are communicating >>> with is not an attacker - then WebID is the right solution. The linkability properties >>> of WebID becomes a positive without the negative of the unlinkability. >>> >>> Perhaps that is how we can summarise the linkability properties of WebID? >> >> I think the question is whether linkability needs to be an unavoidable >> side effect of identifying yourself to a particular site. Clearly >> there are times when you want to be linked and times when you don't. >> Is it clear that the moment you should decide this is the moment when >> you choose your credential? > > You don't have to do it then. You can go to a web site, create an account there, > then later link them up using stronger identities. There is nothing stopping > this from being possible. WebID is just the protocol you can use when you want > strong cross site identity. It does not have to be the authentication system > to use everywhere: clearly not. It doesn't have to be, clearly, but if you want adoption it also seems clear to me that you need to offer a universal solution. If your "web-scale" solution does not actually work for the whole web, why would I adopt it? > >> I think not - and credentials do exist >> that can be used at multiple sites unlinkably, as well as schemes >> where credentials are per-site and so the question is moot. > > And when you want cross site identity WebID is the tool to use. > > So I think we can make some demos where the following can be done: > > 1. You go to a site it does not ask you for any identity > 2. You can login using a username/password, or some unlinkable system eg: per site certificate > 3. you then decide you trust the site, and authenticate with WebID making it linkable. > > I don't see why that is not feasible, you just need to think semantically about identifiers. > That is: don't tie identity to a literal, allow yourself to use leibniz's law to work with > identities across literal identifiers: > > <http://bblfish.net/#hjs> foaf:mbox <mailto:henry.story@bblfish.net>; > foaf:openid <http://bblfish.net/>; > cert:key [ cert:modulus ...; cert:exponent ... ]. > > Here we have 4 identifiers for the user, each with its verification mechanism. Clearly a site > could start off with > > <user1231> local:cookie "Cookie1231" . > > And later add > > <user1231> local:cookie "Cookie1231" ; > cert:key [ cert:modulus ...; cert:exponent ... ] . > > where it only has a per site key for the user. And later > > <user1231> owl:sameAs <http://bblfish.net/#hjs> ; > local:cookie "Cookie1231" ; > cert:key [ cert:modulus ...; cert:exponent ... ] . > > >> >>> >>> Henry >>> >>> [1] http://tools.ietf.org/html/draft-iab-privacy-terminology-01#section-4 >>> >>> Social Web Architect >>> http://bblfish.net/ >>> > > Social Web Architect > http://bblfish.net/ >
Received on Monday, 8 October 2012 12:05:15 UTC