Message-Id: <199409212309.TAA15474@postman.osf.org>
To: iesg@cnri.reston.va.us
Subject: Re: Last Call: URL to Proposed and URN- and IRL-Reqs to Informational
Date: Wed, 21 Sep 1994 19:09:00 -0400
From: "Norbert Leser - OSF DCE: (617)621-8715" <nl@osf.org>
IESG Secretary:
Regarding your announcement from 12 Sep 94 (Subject: Last Call: URL to Proposed
and URN- and IRL-Reqs to Informational), I herewith want you to consider the
delay of the requested publication of the
"Requirements for Uniform Resource Names"
<draft-ietf-uri-urn-req-00.txt>
draft document as an Informational RFC (currently scheduled for 9/26).
Reasoning:
1. I am aware that the URI working group has already put a vast amount of
effort into the current draft and spent a number of sessions and
discussions in the course of the past two years in order to reach
a consensus and to produce this important document.
However, it is an unfortunate fact that - independently of the URI working
group - an international and multi-company initiative (which I am an active
member of) has been working on the X/Open Federated Naming (XFN)
specification for about the same time.
In analyzing your work and after talking to Karen Sollins and other
experts, I come to the conclusion that there is a fair amount of overlap
in both of our specifications and goals as well as elements in the
specifications that can greatly complement each other.
In order to guarantee that both of our specifications lead
into the same direction, assure cohesiveness, and provide for rapid and
broad acceptance, I view it as essential and urgent that both of us get
together to identify and resolve the issues.
Due to our prior commitment, we did not have the chance to be actively
involved in your work and we have only become recently aware - with
the URN publication on 9/10/94 and the IESG announcement from 12 Sep 94,
that the publication as Informational RFC is scheduled immediately
(9/26/94). This leaves us no time to resolve the issues in a manner that
is satisfying for the community.
You should also be aware that the XFN specification has been accepted by
X/Open in June this year as Preliminary Specification. According to
X/Open's policies, this specification will become an official CAE standard
after an experimental period and further review (typically approx. one year).
This provides us with the unique opportunity to try to align the URN
and the XFN specification work assuming that there is sufficient interest
in the URI working group and the promotion of the current Requirements document
is delayed for a limited period of time. I am committed to actively
participate in this communication (As a matter of fact, I'll be sending a more
detailed analysis of the encountered problems to the uri mailing list
during the next few days).
2. Both the URI (RFC1630) and the "Requirements for Uniform Resource Names"
document contain cross references to each other. The URN document states
that URN names are to "fit within the overall architecture of Uniform
Resource Identification".
While the URI specifies syntactical rules that appear to be suited for
the currently listed URLs, the imposed syntax and encoding restrictions
are not suitable for existing naming systems.
Only after talking to Karen Sollins about the relationship of URN to
the URI document, I found that there might not be such a dependency.
At least, what needs to be modified in the URN document is a precise
definition of what the document dependencies are and non existing
dependencies on RFC 1630 must not be suggested in any way.
3. One of the central claims in the URN document is "the support of
existing legacy naming systems". Some of these envisioned naming
systems are enumerated, namely "ISBN numbers, ISO public identifiers,
UPC product codes". While the requirements for resource names might
be appropriate and sufficient for these particular naming systems,
there is a number of other widely used naming systems that cannot
sufficiently be mapped onto the proposed model. Some of these naming
systems that come to ones mind are the ISO X.500 Directory Service, the
OSF DCE Directory Services, the NetWare Version 4 NDS, NIS+,
and the generic naming interfaces of the recently published XFN
Preliminary Specification from X/Open.
If it is not the intention of the URN requirements specification to
provide support for the mentioned and other broadly deployed naming
systems, the wording in the current document that suggest the
"support of existing legacy naming systems" must be changed
accordingly.
Many of the requirements specified in the "Requirements for Uniform
Resource Names" document do not map well onto the listed and other
naming systems. The most significant are:
- As I discussed in my second point above, the URI specifies
syntactical rules that are not suited for the targeted naming
systems.
- URN names are required to express names with "global scope".
Again, a number of existing legacy naming systems permit
context relative names in order to perform sufficiently
and to hide complexity from naming system users.
- URN names are to be parsed and compared "case insensitive".
Many naming systems require to be case-exact or permit the
case matching rules to be configurable. Users of these naming
systems could not safely utilize URN names.
My request is to delay the consideration of this "Requirements for Uniform
Resource Names" draft specification until listed problems are being addressed
and solved in a manner such that at least the major existing naming systems
can sufficiently be supported.
Even if an Informational RFC doesn't preclude further modifications and
refinements, I consider it important for these deficiencies to be
addressed at this stage of specification development. Further URN
specification work could not sensibly been done without a reasonably
stable requirements document.
This request for dealy is supported by the "champions" responsible
for the X/Open Federated Naming specification, representing the
companies Banyan, Digital, HP, IBM, SNI, SunSoft, and OSF.
Sincerely
Norbert Leser