shibboleth-dev - ARP selection algorithm
Subject: Shibboleth Developers
List archive
- From: Scott Cantor <>
- To:
- Subject: ARP selection algorithm
- Date: Fri, 19 Apr 2002 13:00:02 -0400
- Importance: Normal
- Organization: The Ohio State University
Maybe this is because the discussions have tended to oversimplify some
of the details, but I just want to clarify the way the AA is going to
need to locate those policies. I'm not sure if I'm seeing a problem or
not.
Because of the wildcarding, you can't view the whole thing as strictly a
tree, or a sequence of containers (SHAR, Resource, ...) or else you'll
miss some. That might be why I had to think a while to understand it
that way, since I tend to view things as relational when I'm designing
(so I'm not an OO guy, sue me).
Here's an example set of policies. Doesn't really matter whether they're
admin or user-set.
SHAR: *
URL: *
Attr: affiliation=member
SHAR: *.edu
URL: *
Attr: affiliation (all values)
SHAR: chem101.osu.edu
URL: http://chem101.osu.edu/syllabus/*
Attr: entitlement=urn:mace:osu.edu:courses:chem101
SHAR: chem101.osu.edu
URL: http://chem101.osu.edu/gradebook/*
Attr: entitlement=urn:mace:osu.edu:courses:chem101
EPPN
Now, say I access http://chem101.osu.edu/special/page.html
What gets released? The correct answer is affiliation, based on the
second ARP above.
But you don't get that result if you follow the matching steps in a
strict order, and check the SHAR first, before you check the URL. If you
do that, you throw out the first two and only see the last two, since
they match the SHAR exactly. Then you try the URLs and neither match, so
you end up releasing nothing.
My point is that you can't know whether the ARP matches just by looking
at the SHAR, and since you can't know that, you can't throw any matching
ARPs out for being less-specific matches until you go down a level and
look at the URL.
So the real process is to simultaneously match both SHAR and URL to get
a matching set, and *then* you pick the best match for SHAR. That would
give you the proper result above, since the matching set ends up being
the first two, and you pick the second as a better match.
Maybe this is already understood, but I wanted to make sure. I'm
changing the AA section in the architecture to better explain all this,
because I took out some of Marlena's verbose examples that did this and
took out too much in the process.
-- Scott
------------------------------------------------------mace-shib-design-+
For list utilities, archives, subscribe, unsubscribe, etc. please visit the
ListProc web interface at
http://archives.internet2.edu/
------------------------------------------------------mace-shib-design--
- ARP selection algorithm, Scott Cantor, 04/19/2002
- Re: ARP selection algorithm, Parviz Dousti, 04/19/2002
- RE: ARP selection algorithm, Scott Cantor, 04/19/2002
- Re: ARP selection algorithm, Parviz Dousti, 04/19/2002
Archive powered by MHonArc 2.6.16.