John Merrells wrote:
>> Different names? So, are you saying the sxip#1 capabilities only come from sxip and if >> I want to supply say, a user first name, from some other authority I would need to
>> name the property differently?


With dmd1yes. The query mechanism is super brain dead simple... you can reference a property value and that's it.

So, what is a home site to do if sxip attributes are requested? Run off to sxip.net with an intermediate request (utilizing a previously agreed trust), or send the user as a membersite with a fetch-request, fail the request? Seems to me that the membersite ought to be getting those attributes from sxip.net, who is acting as a home site, nobody else needs to be involved - not even the draft :)


Or we could extend the query language to make this type part of the
request, but I'm trying to avoid yet another query language...

Perhaps this is the missing dix#1 capabilities, but the benefits of this are substantially reduced if instead of identity silos, we end up with attribute silos. The example attributes that are sxip#1 in that document are prime candidates for dix#1, and sxip#1 if anything, should be a separate draft detailing that schema - and btw, where do the dix#1 capabilities fit into this model, who supplies those, or is dix#1 the (to come) common schema as I am hoping?

I have to say I am not sold on the idea of compartmentalizing properties first by source. I would rather see a property type schema where the source of the property is not tied down, only its meaning. The source of the property can be disclosed (as a signature) for those membersites that care, then user supplied attributes may simply not be signed. In fact, in many common cases the source _will_ be the user anyway.

dix use case:

About 10 years ago I realized amazon basically had the monopoly on my book buying, then video buying, then whatever because the barrier for me was filling out another form with a bunch of details I don't have memorized. Nothing has changed in those 10 years. So, I want to go to Barnes and Noble and buy a book without filling out any forms, and without "joining" the website in any way, I want a one time transaction just like I get in a store, and given that I am authenticated with my homesite I want a one click transaction. All details supplied are no more or less trustworthy than those I would type in myself even if my browser were acting as the homesite.

This doesn't work if I have amazon#1 properties and B&N want B&N#1 properties. It should be hard for that situation to arise, and the least path of resistance should be an ietf draft detailing the generic schema.

--

Pete

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

_______________________________________________
dix mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/dix

Reply via email to