Re: Stable URL's for working resources?

From: Geoffrey M. Clemm (geoffrey.clemm@rational.com)
Date: Thu, Apr 06 2000

  • Next message: Geoffrey M. Clemm: "Re: Stable URL's for working resources?"

    Date: Thu, 6 Apr 2000 17:59:00 -0400 (EDT)
    Message-Id: <200004062159.RAA04058@tantalum.atria.com>
    From: "Geoffrey M. Clemm" <geoffrey.clemm@rational.com>
    To: ietf-dav-versioning@w3.org
    Subject: Re: Stable URL's for working resources?
    
    
       From Geoff Clemm
    
       I'd like to remove the requirement that a workspace resource
       provide stable URL's for its working resources.
    
       The problem is that many/most implementations don't store the
       working resources in a repository, but rather in a simple file
       system tree.  So although it is reasonable to expect an
       implementation to maintain stable URL's for the versioning
       metadata such as revisions and activities, it is not as reasonable
       to expect it to maintain them for working resources. Since the
       whole point of a working resource is for it to be mutable,
       requiring it to be identifiable via an immutable URL is not what
       one would expect in any case.
    
       The only change that this requires to the protocol is to change the
       definition of the DAV:working-resources property of a workspace
       resource to no longer require that the href's contain "stable
       URL's".
    
    
       From: jamsden@us.ibm.com
    
       I'm OK with this, but workspaces should not be concerned with
       implementations that store resources locally in the client file
       system.  This is purely a client issue for optimizing access,
       supporting file-based access to resources, and supporting
       disconnected work.
    
    I wasn't referring to storing resources locally in the client
    file system (I agree that is out of scope of the protocol), but
    rather for servers that are designed to support thousands of
    workspaces, and do so by maintaining a server farm, each of which
    contains some number of workspaces.  The "file system tree" was
    a reference to a common server side implementation technique.
    Since many server extensions require the instantiation of resources
    in a file system tree, the file system implementation is a key one.
    
    Even for non-file system implementations, the general principle of
    decoupling a workspace as much as possible from the shared versioning
    repository will be a key element of scaling and performance.
    
    Cheers,
    Geoff