[OGSA-AUTHZ] checkpointing the discussion on VO attributes

[Valerio Venturi]

Hi,



>> > > Sounds good. Another similar issue has been issued here an there,
>> > > defining XACML attributes and obligation needed for authorization
>> > > services. What about including that too? This is something that those
>> > > implenting authorization services are facing, as you know, and community
>> > > consensus would be very important.
>> > > Also, deciding that may help in sorting out one of the main concern with
>> > > the current authz decision spec, that is, having or not having attribute
>> > > and obligation definition in the profile.
>> > > If we can be sure to have those defined in a separate document to be
>> > > released soon, may be it's ok to remove them from the current spec.
>> > > DavidC, what do you think about that?
>>     
> > 
> > As I have said all along, I think defining attributes and obligations is 
> > a long term project that will mature as more people start to use them. I 
> > dont think a quick fix spec is the correct approach, because if it is 
> > quick, it wont be complete, and if it is complete it cannot be quick.
>   
You're totally right, a quick fix spec is not what we need. Having
a new group working on that seems to me going in the direction
of more thought, complete documents, which is what we all would
prefer.

> > Therefore the approach that I have been advocating is a two step one. A 
> > quick first stab at a few core attributes and obligations (either in an 
> > existing doc so that a charter change is not needed) or as a separate 
> > doc (in which case a charter change or new WG is needed) - I dont 
>   
Honestly I've a mild opinion on that. Probably it would be cleaner to
have separated docs, as someone in the WG suggested, but I don't have
problems with having some attributes defined in the authz decision
request spec, and definitely this is not a showstopper.

> > actually mind which approach is taken. But I dont support the creation 
> > of a new WG, since this will only dilute the effort we have, and it is 
>   
I understood that the OGSA AuthZ WG is going to finish its work shortly
after next OGF, or at least continuining at a minimum level, 
having the chartered docs in public comments. So I thought there wouldn't 
have been a big overlapping between the two WGs.
Am I wrong?
I implicitly suggested a change to the charter to include the attribute
doc in my first mail, and my impression from talking with DavidG and
Blair in the last momths is that they don't fell like changing the WG
charter.
Probably a clarification from the area directors on that is needed,
since this is also related on how they see the future of the security
area.


> > likely to grow to include further topics (such as obligations  :-)  as one 
> > sees fit. If one is concerned about the progress of the current set of 
> > Authz documents it is because very few people are actually contributing 
> > to them, and some that are working in the area do not wish to actively 
> > contribute.
>   
I see your point, and I share your disappointment. But we cannot go
chasing people and force them to commit. It's their choice, and either
they're not interested in standards or they think it wasn't worth committing.
As a condition for starting a new WG, we must have a real interest from
the community. This means both time to commit, and implementation
experience to share.

> > After the first quick fix has been published then a much longer term 
> > project to produce a richer set can start. This longer term project must 
> > have an active dynamic set of attributes and obligations that can be 
> > added to as the need arises, rather like an IETF/IANA registration 
> > authority for well known ports. So it might be a web page that publishes 
> > this, rather than a paper document. It musn't be a static set, of that I 
> > am sure (I have enough experience of LDAP schemas to know this)
>   
Sounds good. I see you're the first to have ideas and attention in this
topics. I think those things are worth a community effort, I don't mind
what the name of the group is going to be.

Valerio