Soundness (was Re: [xdi] Minutes: XDI TC Telecon Thursday 1-2:00PMPT 2010-10-21)

["Giovanni Bartolomeo" <giovanni.bartolomeo@uniroma2.it>]

Hello Mike,

thank you for your reply. Please see my answers below.

Kind Regards,
Giovanni

Def. Quota "Michael Schwartz" <mike@gluu.org>:

>
> So the problem child in your diagram is
>  +friend/=alice/+knows ???
>
> Firstly, I would argue that in the diagram you supplied, the root  
> node you identified is compound: its =bob+friend.

uhmm.. this is an important point to clarify. Basically I agree that  
the absolute identifier for that node is =bob+friend. This is useful  
also to avoid confusion between multiple nodes which may be labeled as  
"+friend" in different contexts (=bob+friend, =alice+friend, etc.) -  
pls see also second slide attached to this email.

So I agree that - contrary to the second slide in the attachment -  
there are DIFFERENT nodes, =bob+friend, =alice+friend which represents  
different sets of friends of different people.

Now, the precise form we should use whenever +friend is used as a  
predicate should be then =alice/=alice+friend/=bob. But the issue is  
that we have always written =alice/+friend/=bob. Maybe we can justify  
this saying that to avoid such verbosity, we simply use the form  
=alice/+friend/=bob, with the understanding that, when no other  
direction is given, by default the predicate (+friend) is always  
implicitly referred to the subject (i.e. it should be read  
=alice+friend when it appears in the context of =alice).

What do you think?

>
> Secondly, I see this as parallel to LDAP, where my DN (distinguished  
> name) might be:
>  uid=mike,ou=people,o=gluu
> So if I say uid=mike,ou=people, it is ambiguous. Similarly in HTTP  
> URLs, you have relative and absolute URLs.
>
> Do we need some rules that distinguish relative and aboslute XRI's ?
>

maybe... but I what I'm suggesting is that +friend/=alice/+knows has  
no meaning and must not be inferred from the graph for any reason. Any  
syntactical solution allowing to EXCLUDE that statement is fine with me.

To be very clear, what I do not want at all is that, at the end, this  
statement could be somehow legitimate because you can read it in that  
graph, whereas it was not in the set of original statements, NEITHER  
in the set of statements which could be correctly inferred (e.g. using  
a reasoner) from that set. Note that this is of fundamental importance  
to preserve soundness [1] in XDI.

[1] http://en.wikipedia.org/wiki/Soundness

----------------------------------------------------------------
Invito da parte dell'Ateneo:
Il tuo futuro e quello della Ricerca Scientifica hanno bisogno del
tuo aiuto. Dona il  5 x mille all'Universita' di Roma Tor Vergata
codice fiscale: 80213750583 http://5x1000.uniroma2.it

concerns-v2.pdf