- From: Jon Radoff <jradoff@novalink.com>
- Date: Thu, 01 May 1997 14:42:57 -0700
- To: -=jack=- <jack@twaxx.twaxx.com>
- CC: "Ron Daniel, Jr." <rdaniel@lanl.gov>, w3c-dist-auth@w3.org
I will put together what I've worked on and make it available via ftp
for anyone to take a look at within a day or so.
Here is the outline of the concept I had in mind:
1. Define an API which would exist in a shared-library type space on
the server (or a DLL on NT).
2. Applications that wanted to be able to verify if a user has a
certain permission would make API calls to do so and respond
accordingly.
3. The shared-library containing API calls would be able to connect
to compliant modules defined by the system administrator (e.g.,
if vendor X wanted to provide a module that makes them
compatible with the API, they'd ship this as a component -- not
unlike how ODBC works in the database world...)
4. A basic concept of the API is to abstract the concept of
authentication and let the application worry about this (it
may be that we want to think of an interface specification for
authentication data too, but it wasn't part of my original idea).
We should discuss the pros/cons of this.
5. The API attempts to give the concepts around security a
"real-world" feel. Users own abstract, named permission
entities as opposed to traditional read/write/execute
permissions. That way, applications and security management
systems are free to define what a given named permission
entity means.
Other items for discussion:
a. We should discuss whether it makes sense to include in any
standard the ability to define new permission entities; I was
leaning against this because I thought if we didn't keep it
abstract it could limit the creativity of what the "permission
server" vendors.
b. Should the abilty to assign permission entities to a content
object be standards defined, or application defined? I think
we probably need a combination of (i) basic permissions which
would exist in every DAV-compliant application and
(ii) the ability for applications to form their own rules
based on permission entities they could provide. What about
permissions on an object that are more granular than overall
access to the URI (such as a piece of a particular page?)
c. Does this track make sense at all? ;)
I will let everyone know when my stuff is available...
Jon
Received on Thursday, 1 May 1997 14:42:10 UTC