RE: Issue: Requiring server to use / terminated URL for returned collections

   From: Lisa Dusseault []

   > ... I need to see some compelling reasons why a
   > collection requires protocol-imposed naming constraints.

   Another reason was given by Roy in an older mail:

   "There are no known examples of HTTP servers for which the URI
   without a slash is the SAME RESOURCE as the URI with a slash.

That is because pre-WebDAV HTTP servers just redirected a request on
what WebDAV identifies as a collection (with or without the last
slash) to some server-defined member of that collection
(e.g. index.html).

   Automatically providing two different URI for the same resource,
   particularly when they cross hierarchical syntax, instantly doubles
   the application name space (bad for IR and QA) and creates holes in
   access control."

If that is in fact a problem, then the server won't do it.  On the
other hand, if the server's implementation does not support trailing
separator characters, and the user expects this equivalence because
all other protocols that access this server do so, it may be
significantly more complex (for both the implementor and a user of the
repository) to try to fake trailing separator characters as indicating
a separate resource.  That is why the mapping of URIs to types of
resources needs to be under the server's control, not predefined by
the protocol.

   If the server consistently uses the same URL for each resource,
   then it's easier for clients and intermediaries to handle (cache,
   compare, concatenate paths, etc) those resources.

With the presence of multiple bindings to the same resource
(something that is explicitly allowed by the rfc2616, and
explicitly supported by the BIND protocol extension), the
client cannot assume that there is only one URI for a given
resource.  Any client that assumes otherwise has a serious bug
that should be fixed.

   > Since when is a collection path without a trailing slash a "bad
   > Request-URI"?  RFC-2518 explicitly states in Section 5.2 that "a
   > resource may accept a URI without a trailing '/' to point to a
   > collection."  If you are proposing to change this semantics, then
   > I object even more vigorously (but I'll save those objections
   > until I verify that you are proposing to make this change).

   I don't think it matters too much which one you pick, with trailing
   slash or without (with trailing slash is slightly more convenient
   because it's a marker for collections). However, there was strong
   opinion that the spec should pick one. Allowing such flexibility,
   as usual, makes life harder for clients.

The only client complexity that we have identified is the need for the
client to add a '/' before adding a new segment, if there is not yet a
slash there (usually in one line of code).  Given that file system
clients have dealt with this for years with no issues, I see no reason
why this is any harder for HTTP/WebDAV clients.  If the spec is
ambiguous, or the algorithms required are complex, then certainly one
can fix the protocol, but if a client just has a bug with a simple
fix, you fix the client -- you don't change the protocol to workaround
that bug for a few special cases.

   > Note also that there is no requirement in 2518 for a server to
   > redirect a non-GET request (it just needs to return the '/'
   > terminated name in a Location header).  Note that I have no
   > problem with requiring that a Location header be returned, since
   > if the server is applying a method to a resource, it will have
   > had to find out what type of resource it is anyway.

   Yes. And if the server returns the slash-terminated name in a
   Location header, it should use the SAME name everywhere else in the

That is an opinion with which I disagree.

   > The only real interoperability problem I saw identified above was
   > the "clients don't redirect properly", and the solution there is
   > to fix the broken clients, not change URL conventions so that the
   > broken clients work better in this one case of redirection.  (I
   > consider the addition of a slash when extending a URL to be a
   > trivial coding issue, not an interoperability problem).

   Also an interoperability problem found was "clients don't append new
   child names properly onto collection names".

That was the "trivial coding issue" I was referring to.  A client that
appends a new segment to a URL without ensuring there is a
preceding '/' just has a bug, and a bug that is trivial to fix.


Received on Wednesday, 18 September 2002 10:39:23 UTC