- From: Daniel Brotsky <dbrotsky@adobe.com>
- Date: Wed, 20 Mar 2002 23:49:01 +0100
- To: "Lisa Dusseault" <ldusseault@xythos.com>
- Cc: <w3c-dist-auth@w3.org>
Great find, Lisa. This is an issue we've been starting to focus on at
Adobe, as well. I would be willing to collaborate on defining a WebDAV
extension for SLP Service Agents (and interested clients).
dan
On Wednesday, March 20, 2002, at 10:34 PM, Lisa Dusseault wrote:
>
> I've been looking into the problem of locating WebDAV servers on an
> intranet. Companies that have WebDAV servers on their net may not
> wish to
> have users manually configure various clients with the names of all
> those
> servers. In researching this problem, I've found only one
> standards-based
> approach, and that's SLP.
>
> Who would be interested in defining a WebDAV extension so that Service
> Location Protocol (SLP) clients can distinguish WebDAV servers from
> plain
> Web servers? Any feedback on how this should work? Or are there other
> options I'm ignorant of?
>
> The SLP specification: http://www.ietf.org/rfc/rfc2608.txt
> SLP home page, links to open source project: http://www.srvloc.org/
> SLP and the Macintosh: http://www.opendoor.com/shareway/SLP.html
>
> Some background...
>
> SLP models
>
> SLP works in one of two models. In the direct client-to-server model,
> the
> client multicasts requests for a particular kind of service to the local
> network. Services of that type respond directly (unicast) to the
> client. I
> see problems with this model because it would require a WebDAV server to
> receive multicast.
>
> The second SLP model involves a directory agent to register services:
>
> +-------+ -Unicast SrvRqst-> +-----------+ <-Unicast SrvReg- +--------+
> | User | | Directory | |Service |
> | Agent | | Agent | | Agent |
> +-------+ <-Unicast SrvRply- +-----------+ -Unicast SrvAck-> +--------+
>
> This model seems easier for WebDAV servers to implement (they just have
> to
> send their registration and accept the response).
>
> Service URLs
>
> Services advertise themselves using service names or schemes. As HTTP
> servers, WebDAV servers would probably therefore advertise themselves
> with
> the 'http' service type. E.g.
>
> service:http://myserver.com
>
> A client request for "service:http" would match against all HTTP and
> WebDAV
> servers advertising themselves in this way. For all matches, a list of
> attributes for the service can be returned to the client. I suspect the
> attribute list is the right place to advertise WebDAV support. This
> optimizes in a few directions:
> - to find all HTTP servers, whether WebDAV or not, the client only
> needs to
> send one service match request.
> - to find all WebDAV servers, the client only needs to send one
> request,
> and has to filter out the non-WebDAV responses.
> - It is not necessary for the client to send OPTIONS to all HTTP
> servers to
> discover WebDAV support. However, to discover additional WebDAV
> functionality that may be required unless we add additional detail.
> - HTTP/WebDAV servers only need to advertise themselves once, as a http
> service with WebDAV support attributes.
>
> This approach follows the approach for printers, which you can follow
> through examples in the RFC.
>
> Lisa
>
Received on Wednesday, 20 March 2002 17:49:21 UTC