- From: Marc Slemko <marcs@znep.com>
- Date: Thu, 25 Oct 2001 00:02:36 -0700 (PDT)
- To: http-wg@cuckoo.hpl.hp.com
Not really ontopic since it has nothing to do with any HTTP standards, but in terms of real world behaviour... just a FYI in case people start seeing odd things as a result of this. IE6 has a new "passport authentication" scheme built in (same level as basic, digest, etc.). While IE6 has been around for a while, MS is only just now (night before XP launch, of course) deploying upgraded passport servers that use it. It looks to be somewhat of a perversion of the spec (sure, no matter how they do it, it isn't "per spec" since the spec doesn't cover such a thing, but some ways completely violate the whole concept of HTTP more than others), but that isn't new. The server sends something like: WWW-Authenticate: Passport1.4 lc=1033,id=3,ru=http://registernet.passport.com/mobilepin.srf%3Fru%3Dhttp://memberservices.passport.com/memberservice.srf%253Flc%253D1033%26lc%3D1033%26cbid%3D486%26id%3D3%26cb%3D%26lc%3D1033,tw=20,kv=2,ct=1003992089,cb=,cbid=486,ems=1,ver=2.1.0068.1,tpf=XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX,cbid=486 to request authentication...only it doesn't send it with a 401, but with a 302, presumably so non-IE6 browsers will ignore it and get the "normal" login swizzling process. Then the browser magically figures out to authenticate (built into XP, of course) and sends a "normal" authorization header: Authorization: Passport1.4 from-PP='t=XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX' (identifiers X'ed out) The odd thing I have seen, is that if I authenticate in one browser window, then open a second window, I am not always authenticated in the second browser window. That smells of some really ugly violation of the concept of how HTTP is "supposed to work". I'd be quite interested if anyone has any pointers to docs on the protocol, nothing obvious on MS's site aside from a few references to it. I certainly agree that nothing out there really satisfies the most important needs for authentication on the web, but am far from convinced that this MS attempt adds any value at all. Anyway, just FYI. God, what an awful ugly massive security hole MS's passport implementation is. But this isn't the place for that rant...
Received on Thursday, 25 October 2001 08:04:33 UTC