To: uri@bunyip.com
Subject: comments on naming from Steve Strasen
From: Larry Masinter <masinter@parc.xerox.com>
Message-Id: <94Jun11.175801pdt.2760@golden.parc.xerox.com>
Date: Sat, 11 Jun 1994 17:57:47 PDT
Steve Strasen has worked on various Xerox and ISO standards that include naming
mechanisms. Here's his comments on the URN requirements document:
================================================================
The referenced document makes a few very important points which I think are central to this discussion.
1. Names are not the same as addresses (locations).
Perhaps because this [discussion] has been concerned with the assignment of identifiers to network entities, much of the discussion has focused on naming schemes which provide information as to the network location of these entities, i.e. an address of some form or another. Veriy interesting, but not the same as naming.
2. The name is not the same as that which is named.
Rules for name equality are independent of rules for equality of named entities. Names should not be expected to convey information about the characteristics of that which is named. Things named by different names are presumptively but not necessarily different (aliases are allowed).
3. The naming scheme is not dictated by the name service characteristics.
The purpose of the naming scheme is to provide a mechanism for the assignment of unambiguous names as identifiers. It must be independent of the service(es) which provide name-to-characteristics or name-to-location lookup. That is, not only should the name not be expected to specify the location of the named object, it should not be expected to specify the location of a service which can provide the characteristics or the location of the object.
NAMING SCHEMES
The only naming schemes which I know of that are sufficiently general, extensible, and usable are based on the following:
a. The name syntax consists of a sequence of one or more name-components.
b. The conditions under which two name-components are equal is defined by the naming scheme. Two names are equal if and only if they have the corresponding name components are equal.
c. There is a set of root naming authorities defined by the naming scheme itself or registered in a root registry. The first name-component of any name identifies a root naming authority.
d. Each naming authority owns the name space of all names which begin with a specified sequence of name-components, and has the right to assign any names in that name space and/or to assign a portion of its name space to a sub-authority.
NOTES -
i. Delegation of a naming authority does not imply definition of the entities named.
The delegation of naming authority does not require any definition of the characteristics of the entities named under the delegated authority.
ii. Registration of a name assignment does not imply registration of the entity named by that name or of any of the characteristics of that entity (with the possible exception of the identity of its defining authority).
EXISTING NAMING SCHEMES
1. Interpress universal names
The Interpress "universal name" scheme meets all of the requirements for naming, but it remains a proprietary scheme and it has not been used extensively. Also, Interpress universal names are intended to be unique only across a specified class of objects, which could cause problems in an environment in which the object classes are not well defined.
2. ASN.1 Object Identifiers
ASN.1 object identifiers (defined by ISO/IEC 8824) are badly underspecified and have several practical difficulties, but are widely used in international standards and OSI environments. The only representation defined (by ISO/IEC 8825) requires a binary communications environment (i.e. is not compatible with clear-text-only transport mechanisms).
3. ISO 9070 public identifiers
ISO/IEC 9070 is identified as a defining an "SGML Support Facility", but it defines a naming scheme which is broadly usable. It is not so broadly implemented as ASN.1 object identifiers and forces a syntactic distinction between "owner identifier" part and "object identifier" parts.
....