http://depts.washington.edu/ontheroa/?p=506 The UW blog for all things Resource Oriented Architecture & web services - email appdev@u.washington.edu Tue, 07 Feb 2012 00:23:05 +0000 hourly 1 http://wordpress.org/?v=3.0 http://depts.washington.edu/ontheroa/?p=506&cpage=1#comment-401 Scott Stephenson Thu, 22 Jul 2010 17:56:42 +0000 http://depts.washington.edu/ontheroa/?p=506#comment-401 I neglected to note that if the payload changes based on authz -- even with the per-attribute 403's -- it makes caching more challenging, less effective and riskier. We'd need to cache a page for each authz role and make sure we didn't accidentally give the wrong one out (a new attack vector). I neglected to note that if the payload changes based on authz — even with the per-attribute 403′s — it makes caching more challenging, less effective and riskier. We’d need to cache a page for each authz role and make sure we didn’t accidentally give the wrong one out (a new attack vector).

]]> http://depts.washington.edu/ontheroa/?p=506&cpage=1#comment-400 Gary Prohaska Wed, 21 Jul 2010 20:01:34 +0000 http://depts.washington.edu/ontheroa/?p=506#comment-400 Regarding the final question for isomorphic payloads, the security concern is somewhat ameliorated by fact that end user may not be aware of disallowed fields if application does due diligence in hiding them -- maybe part of standard SLA for app clients? Regarding the final question for isomorphic payloads, the security concern is somewhat ameliorated by fact that end user may not be aware of disallowed fields if application does due diligence in hiding them — maybe part of standard SLA for app clients?

]]> http://depts.washington.edu/ontheroa/?p=506&cpage=1#comment-399 Scott Stephenson Wed, 21 Jul 2010 19:32:04 +0000 http://depts.washington.edu/ontheroa/?p=506#comment-399 I prefer the idea of an isomorphic payload and am fine with the idea of a per-attribute 403. Furthermore if the attribute includes the URI of another resource, it seems like the URI should still be included even if the client isn't authorized to view it since that's taken care of by HTTP. I prefer the idea of an isomorphic payload and am fine with the idea of a per-attribute 403. Furthermore if the attribute includes the URI of another resource, it seems like the URI should still be included even if the client isn’t authorized to view it since that’s taken care of by HTTP.

]]>