Stable URLs

From: Tim Ellison/OTT/OTI (Tim_Ellison@oti.com)
Date: Wed, Apr 05 2000

  • Next message: Geoffrey M. Clemm: "Re: Stable URLs"

    To: ietf-dav-versioning@w3.org
    Message-ID: <OFDCE80EBC.AE5EF45D-ON852568B8.0066D91F@ott.oti.com>
    From: "Tim Ellison/OTT/OTI" <Tim_Ellison@oti.com>
    Date: Wed, 5 Apr 2000 15:24:33 -0400
    Subject: Stable URLs
    
    Here's my proposal for stable URLs.
    
    Assumptions:
        (1) A URL that is not "stable", is "dynamic".
        (2) Stable URLs are server specific, and cannot be meaningfully parsed
    by the client (i.e., reverse engineered into component parts).  They can be
    thought of as opaque tokens that conform to URL rules so that they can be
    passed as request URIs.
        (3) Stable URLs are indistinguishable from dynamic URLs.  That is,
    there is no mangling of URLs to indicate its stability.
        (4) There is no visible 'meta' area of a server URL namespace.  The
    stable URL space is the exclusive domain of the server.
    
    Axioms:
    - The stable URl forms the equivalent of a server specific unique identity
    of a resource.
    - All resources, (revisions, non-versioned resources, working resources,
    ...) have a stable URL.
    - A resource may be reached by zero or more dynamic URLs.
    - There is a 1:1 mapping from resource to stable URL.
    - Clients can determine the stable URL of any resource they can 'reach' by
    stable or dynamic URL.
    - Methods have the same effect if applied to a resource via its dynamic or
    stable URL.
    
    Usage:
    Since URLs can be dynamic or stable for any request to the server, there
    must be some indication of its stability in the request.  The stability of
    URLs in the response is defined in the protocol specification.  For
    example, some properties are defined as containing stable URLs.  Any
    request that uses a stable request URI must contains <href> elements that
    are themselves stable URLs.
    
    The general form of a request is as follows:
         METHOD <request-uri> HTTP/1.1
         Workspace: [stable] <workspace-url>
         Target-Selector: <keyword> [<param>]
    
    where <keyword> <param> pairs may be one of:
    
    _Unspecified_
         The request-uri is a dynamic URL.  Select the resource reached
         by resolving the request-uri in the context of the request workspace.
         If the selected resource is a versioned resource, select a revision
         of the versioned resource in the context of the request workspace.
    
    label "my label"
         The request-uri is a dynamic URL.  Select the revision of the
         versioned resource labelled "my label".   Select nothing if there
         is no such labelled revision.  Ignore this header if the target
         resource is unversioned.
    
    revid "rev12"
         The request-uri is a dynamic URL.  Select the revision of the
         versioned resource with revision id "rev12".  Select nothing if
         there is no such revision id.  Ignore this header if the target
    resource
         is unversioned.
    
    metadata
         The request-uri is a dynamic URL.  Select the versioned resource
         itself rather than any revision of it.  Return a bad request if the
    target
         resource is unversioned.
    
    stable
         The request-uri is a stable URL.  Select the revision of the versioned
         resource at the request-uri.
    
    stable metadata
         The request-uri is a stable URL.  Select the versioned resource
         itself rather than any revision of it.  Return a bad request if the
    target
         resource is unversioned.
    
    
    Problems:
    Since discovering the members of a resource is a 'side-effect' of doing a
    PROPFIND depth one query, there is no way to find the members of a
    collection given a stable URL to that collection, since a PROPFIND using a
    stable URL woud return the stable URLs of the members (which cannot be
    parsed to reveal their names).
    
    
    Comments?
    Tim