Network Working Group C. Elliott Internet-Draft Cisco Systems Expires: December 11, 2001 D. Harrington Enterasys Networks J. Jason Intel Corporation J. Schoenwaelder F. Strauss TU Braunschweig W. Weiss Ellacoya Networks June 12, 2001 SMIng Requirements draft-ietf-sming-reqs-02 Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This Internet-Draft will expire on December 11, 2001. Copyright Notice Copyright (C) The Internet Society (2001). All Rights Reserved. Abstract This document describes the requirements of a data modeling language, suitable for the modeling of network management constructs, that can be directly mapped into SNMP [1] and COPS-PR [9] protocol PDUs. Elliott, et. al. Expires December 11, 2001 [Page 1] Internet-Draft SMIng Requirements June 2001 Additionally, it is desirable that if possible the language should be able to be translated into SMIv2 [3], [4], [5] and/or SPPI [10]. This document identifies requirements of an updated data modeling language for SNMP and COPS-PR. The purpose of this document is to ensure that subsequent language specification is complete and consistent with the stated requirements. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 5 2. Motivation . . . . . . . . . . . . . . . . . . . . . . . . 5 3. Background . . . . . . . . . . . . . . . . . . . . . . . . 5 4. Specific Requirements for SMIng . . . . . . . . . . . . . 6 4.1 Accepted Requirements . . . . . . . . . . . . . . . . . . 7 4.1.1 The Set of Specification Documents . . . . . . . . . . . . 7 4.1.2 Textual Representation . . . . . . . . . . . . . . . . . . 7 4.1.3 Human Readability . . . . . . . . . . . . . . . . . . . . 7 4.1.4 Machine Readability . . . . . . . . . . . . . . . . . . . 7 4.1.5 Accessibility . . . . . . . . . . . . . . . . . . . . . . 8 4.1.6 Language Extensibility . . . . . . . . . . . . . . . . . . 8 4.1.7 Special Characters in Text . . . . . . . . . . . . . . . . 8 4.1.8 Naming . . . . . . . . . . . . . . . . . . . . . . . . . . 8 4.1.9 Namespace Control . . . . . . . . . . . . . . . . . . . . 9 4.1.10 Modules . . . . . . . . . . . . . . . . . . . . . . . . . 9 4.1.11 Module Conformance . . . . . . . . . . . . . . . . . . . . 9 4.1.12 Arbitrary Unambiguous Identities . . . . . . . . . . . . . 10 4.1.13 Protocol Independence . . . . . . . . . . . . . . . . . . 10 4.1.14 Protocol Mapping . . . . . . . . . . . . . . . . . . . . . 10 4.1.15 Translation to Other Data Definition Languages . . . . . . 10 4.1.16 Base Data Types . . . . . . . . . . . . . . . . . . . . . 11 4.1.17 Enumerations . . . . . . . . . . . . . . . . . . . . . . . 11 4.1.18 Discriminated Unions . . . . . . . . . . . . . . . . . . . 11 4.1.19 Instance Pointers . . . . . . . . . . . . . . . . . . . . 12 4.1.20 Row Pointers . . . . . . . . . . . . . . . . . . . . . . . 12 4.1.21 Constraints on Pointers . . . . . . . . . . . . . . . . . 12 4.1.22 Base Type Set . . . . . . . . . . . . . . . . . . . . . . 12 4.1.23 Extended Data Types . . . . . . . . . . . . . . . . . . . 13 4.1.24 Units and Default Values of Defined Types . . . . . . . . 13 4.1.25 Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . 13 4.1.26 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . 14 4.1.27 Table Existence Relationships . . . . . . . . . . . . . . 14 4.1.28 Table Relationships . . . . . . . . . . . . . . . . . . . 14 4.1.29 Structures . . . . . . . . . . . . . . . . . . . . . . . . 14 4.1.30 Containment . . . . . . . . . . . . . . . . . . . . . . . 15 4.1.31 Single Inheritance . . . . . . . . . . . . . . . . . . . . 15 4.1.32 Events . . . . . . . . . . . . . . . . . . . . . . . . . . 15 4.1.33 Creation/Deletion . . . . . . . . . . . . . . . . . . . . 16 Elliott, et. al. Expires December 11, 2001 [Page 2] Internet-Draft SMIng Requirements June 2001 4.1.34 Range and Size Constraints . . . . . . . . . . . . . . . . 16 4.1.35 Uniqueness . . . . . . . . . . . . . . . . . . . . . . . . 16 4.1.36 Extension Rules . . . . . . . . . . . . . . . . . . . . . 16 4.1.37 Deprecate Use of IMPLIED Keyword . . . . . . . . . . . . . 16 4.1.38 No Redundancy . . . . . . . . . . . . . . . . . . . . . . 17 4.1.39 Compliance and Conformance . . . . . . . . . . . . . . . . 17 4.1.40 Allow Refinement of All Definitions in Conformance Statements . . . . . . . . . . . . . . . . . . . . . . . . 17 4.2 Nice-to-Have Requirements . . . . . . . . . . . . . . . . 18 4.2.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.2.2 Unions . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.2.3 Abstract vs. Concrete Structures . . . . . . . . . . . . . 19 4.2.4 Float Data Types . . . . . . . . . . . . . . . . . . . . . 19 4.2.5 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.3 Rejected Requirements . . . . . . . . . . . . . . . . . . 20 4.3.1 Incomplete Translations . . . . . . . . . . . . . . . . . 20 4.3.2 Instance Naming . . . . . . . . . . . . . . . . . . . . . 20 4.3.3 Attribute Value Constraints . . . . . . . . . . . . . . . 20 4.3.4 Existence Constraints . . . . . . . . . . . . . . . . . . 21 4.3.5 Ordering Constraints . . . . . . . . . . . . . . . . . . . 21 4.3.6 Attribute Transaction Constraints . . . . . . . . . . . . 21 4.3.7 Method Constraints . . . . . . . . . . . . . . . . . . . . 22 4.3.8 Categories . . . . . . . . . . . . . . . . . . . . . . . . 22 4.3.9 Agent Capabilities . . . . . . . . . . . . . . . . . . . . 22 4.3.10 Relationships . . . . . . . . . . . . . . . . . . . . . . 22 4.3.11 Procedures . . . . . . . . . . . . . . . . . . . . . . . . 23 4.3.12 Associations . . . . . . . . . . . . . . . . . . . . . . . 23 4.3.13 Association Cardinalities . . . . . . . . . . . . . . . . 23 4.3.14 Categories of Modules . . . . . . . . . . . . . . . . . . 24 4.3.15 Length of Identifiers . . . . . . . . . . . . . . . . . . 24 4.3.16 Why Are So Many SMIv1/v2 Parsers So Error Tolerant? . . . 24 4.3.17 Core Language Keywords vs. Defined Identifiers . . . . . . 25 4.3.18 Internationalization . . . . . . . . . . . . . . . . . . . 25 4.3.19 Mapping Modules to Files . . . . . . . . . . . . . . . . . 25 4.3.20 Simple Grammar . . . . . . . . . . . . . . . . . . . . . . 26 4.3.21 Place of Module Information . . . . . . . . . . . . . . . 26 4.3.22 Fully Qualified Identifiers . . . . . . . . . . . . . . . 26 4.3.23 Readable Revision Date Representation . . . . . . . . . . 26 4.3.24 Make Status Information Optional . . . . . . . . . . . . . 27 4.3.25 Remove OIDs from the Core Language . . . . . . . . . . . . 27 4.3.26 Module Namespace . . . . . . . . . . . . . . . . . . . . . 28 4.3.27 Hyphens in Identifiers . . . . . . . . . . . . . . . . . . 28 4.3.28 Referencing a Group of Instances of a Structure . . . . . 28 5. Security Considerations . . . . . . . . . . . . . . . . . 29 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . 29 References . . . . . . . . . . . . . . . . . . . . . . . . 29 Authors' Addresses . . . . . . . . . . . . . . . . . . . . 30 A. Mailing List Discussions . . . . . . . . . . . . . . . . . 32 Elliott, et. al. Expires December 11, 2001 [Page 3] Internet-Draft SMIng Requirements June 2001 Full Copyright Statement . . . . . . . . . . . . . . . . . 42 Elliott, et. al. Expires December 11, 2001 [Page 4] Internet-Draft SMIng Requirements June 2001 1. Introduction This document describes the requirements for the definition of a new object-oriented, data modeling language that can be mapped directly into SNMP and COPS-PR protocol PDUs. It may also be translated into SMIv2 MIBs and SPPI PIBs. Concepts such as structures, attributes, methods, conventions for organization into reusable data structures, and mechanisms for representing relationships are discussed. Conventions used in this document: 2. Motivation As networking technology has evolved, a diverse set of technologies has been deployed to manage the resulting products. These vary from Web based products, to standard management protocols and text scripts. The underlying systems to be manipulated are represented in varying ways including implicitly in the system programming, via proprietary data descriptions, or with standardized descriptions using a range of technologies including MIBs [6], PIBs [11], or LDAP [7] schemas. The result is that network applications and services such as DHCP or Differentiated Services may be represented in many different inconsistent fashions. The SMIng working group will develop a new modeling language to align the languages defined in the SMIv2 and SPPI documents (the languages for writing MIBs and PIBs), since these are very similar. Another motivation is to permit a more expressive and complete representation of the modeled information. Examples of additional expressiveness and completeness that are considered are the ability to clearly define relationships between objects, the expression of constraints on objects and properties, and the ability to define methods. These additional features are discussed in subsequent sections. 3. Background The Network Management Research Group (NMRG) of the Internet Research Task Force (IRTF) has researched the issues of creating a protocol- independent data modeling language that could be used by multiple protocols. Because SMIv2 and SPPI are very similar, the NMRG focused on merging these two languages, but also researched ways to abstract the requirements to produce a language that could be used for other protocols, such as LDAP and Diameter. The NMRG has published the results of their work in [12], and has submitted their specification as one proposal to consider in the development of the SMIng language. Elliott, et. al. Expires December 11, 2001 [Page 5] Internet-Draft SMIng Requirements June 2001 The SMIng Working Group has accepted their submission for consideration, and to use their proposal to better understand the requirements and possible obstacles to be overcome. Where useful, the NMRG proposal has been referenced in the details below. 4. Specific Requirements for SMIng The following sections define the requirements for the definition of an object-oriented, data-modeling language. The draft captures the results of the working group discussions regarding the SMIng requirements process. It is broken up into three sections: accepted requirements (Section 4.1), nice-to-have requirements (Section 4.2), and rejected requirements (Section 4.3). Appendix A contains the requirements discussion that was generated on the SMIng mailing list. Each requirement has the following information: o Number: the original requirement number (as a means for cross referencing) o Type: a field that identifies the type of requirement, using one of the following values: * basic: considered a basic requirement for SMIng and is contained in SMIv2 and/or SPPI. * align: supported in different ways in SMIv2 and SPPI and they must be aligned. * must: considered a fix for a known problem in SMIv2 and/or SPPI. * should: modifies something that is often misused, or would be nice to have if it can be easily done and does not cause additional complexity or delay. * new: considered a new feature that is not required in SMIng, but could be added if working group consensus to do so is reached. o From: a field that defines the origin of the requirement and that contains one or more of the following values: * SMI: exists in SMIv2. * SPPI: exists in SPPI. * NMRG: exists in the current NMRG specification proposal, but not in SMIv2 or SPPI. * Charter: exists in working group charter. * WG: proposed during working group discussions. * Individual: proposed by working group participant. o Description: a quick description of the requirement. o Motivation: rationale for the requirement. o Notes: optional notes about a requirement. For example, for nice- to-have or rejected this may contain reasoning why this requirement is not required by the SMIng working group, but justification why it should be considered anyway. Notes may be the opinions of the requirements process participants and as such should not be taken as consensus of the working group or the Elliott, et. al. Expires December 11, 2001 [Page 6] Internet-Draft SMIng Requirements June 2001 recommendation of the requirements editing team. 4.1 Accepted Requirements This section represents the list of requirements that have been accepted by the SMIng working group. 4.1.1 The Set of Specification Documents Number: 73 Type: new From: NMRG Description: SMIv2 is defined in three documents, based on an obsolete ITU ASN.1 specification. SPPI is defined in one document, based on SMIv2. The core of SMIng should de defined in one document and must be independent of external specifications. Motivation: Self-containment. 4.1.2 Textual Representation Number: 1 Type: basic From: SMI, SPPI, WG Description: SMIng definitions must be represented in a textual format. Motivation: General IETF consensus. 4.1.3 Human Readability Number: 2 Type: basic From: WG Description: The syntax must make it easy for humans to directly read and write SMIng modules. It should be possible for SMIng module authors to produce SMIng modules with text editing tools. Motivation: The syntax should make it easy for humans to read and write SMIng modules. 4.1.4 Machine Readability Number: 3 Type: basic From: SMI, SPPI Description: The syntax should make it easy to implement parsers. A complete ABNF specification of the grammar is desirable. Furthermore, the language should forbid things like forward references unless they are unavoidable. Motivation: A complete specification of the language grammar in ABNF Elliott, et. al. Expires December 11, 2001 [Page 7] Internet-Draft SMIng Requirements June 2001 encourages the use of compiler toolkits to construct solid parsers. Avoiding unnecessary forward references simplifies parser internal data management and allows for early error detection. 4.1.5 Accessibility Number: 17 Type: align From: SMI, SPPI Description: Attribute definitions must indicate whether attributes can be read, written, created, deleted, and whether they are accessible for notifications, or are non- accessible. Align PIB- ACCESS and MAX-ACCESS, and PIB-MIN-ACCESS and MIN-ACCESS. Motivation: Alignment. 4.1.6 Language Extensibility Number: 52 Type: new From: NMRG Description: The language should have characteristics, so that future modules can contain information of future syntax without breaking original SMIng parsers. E.g., when SMIv2 introduced REFERENCEs it would have been nice if it would not have broken SMIv1 parsers. Motivation: Achieve language extensibility without breaking core compatibility. 4.1.7 Special Characters in Text Number: 57 Type: new From: Individual Description: Allow an escaping mechanism to encode special characters, e.g. double quotes and new-line characters, in text such as DESCRIPTIONs or REFERENCEs. Motivation: ABNF can contain literal characters enclosed in double quotes; to provide the ABNF grammar, there must be the ability to escape special characters. 4.1.8 Naming Number: 4 & 5 combined Type: basic From: SMI, SPPI Description: SMIng must provide mechanisms to uniquely identify Elliott, et. al. Expires December 11, 2001 [Page 8] Internet-Draft SMIng Requirements June 2001 attributes, groups of attributes, and events. It is necessary to specify how name collisions are handled. Motivation: 4.1.9 Namespace Control Number: 5 Type: basic From: SMI, SPPI Description: There should be a hierarchical, centrally-controlled namespace for standard named items, and a distributed namespace should be supported to allow vendor-specific naming and to assure unique module names across vendors and organizations. Motivation: Need to unambiguously identify definitions of various kinds. Some SMI implementations have problems with different objects from multiple modules but with the same name. Furthermore, the probability of module name clashes rises over time (for example, different vendors defining their own SYSTEM- MIB). Notes: An example naming scheme is the one employed by the Java programming language with a central naming authority assigning the top-level names. The working group has to make a determination as to how best to handle namespace control (e.g. a BCP). 4.1.10 Modules Number: 6 Type: basic From: SMI, SPPI Description: SMIng must provide a mechanism for uniquely identifying a module, and specifying the status, contact person, revision information, and the purpose of a module. SMIng must provide mechanisms to group definitions into modules and it must provide rules for referencing definitions from other modules Motivation: Modularity and independent advancement of documents. Notes: Text about module conformance has been moved to Section 4.1.11. 4.1.11 Module Conformance Number: 6 Type: basic From: SMI, SPPI Description: SMIng must provide mechanisms to detail the minimum Elliott, et. al. Expires December 11, 2001 [Page 9] Internet-Draft SMIng Requirements June 2001 requirements implementers must meet to claim conformance to a standard based on the module. Motivation: Ability to convey conformance requirements. 4.1.12 Arbitrary Unambiguous Identities Number: 66 Type: basic From: SMI Description: SMI allowed the use of OBJECT-IDENTITIES to define unambiguous identities without the need of a central registry. SMI uses OIDs to represent values that represent references to such identities. SMIng needs a similar mechanism (a statement to register identities, and a base type to represent values). Motivation: SMI Compatibility. Notes: This is an obvious requirement. Additionally, everything not on the wire, such as modules, will still be assigned OIDs. It is yet to be determined whether the assignment of the OID occurs within the core or within a protocol-specific mapping. 4.1.13 Protocol Independence Number: 7 Type: basic From: Charter Description: SMIng must define data definitions in support of the SNMP and COPS-PR protocols. SMIng may define data definitions in support of other protocols. Motivation: So data definitions may be used with multiple protocols. 4.1.14 Protocol Mapping Number: 8 Type: basic From: Charter Description: The SMIng working group, in accordance with the working group charter, will define mappings of protocol independent data definitions to protocols based upon installed implementations. The SMIng working group can define mappings to other protocols as long as this does not impede the progress on other requirements. Motivation: SMIng working group charter. 4.1.15 Translation to Other Data Definition Languages Number: 9 Type: basic From: Charter Elliott, et. al. Expires December 11, 2001 [Page 10] Internet-Draft SMIng Requirements June 2001 Description: SMIng language constructs should, wherever possible, be translatable to SMIv2 and SPPI. At the time of standardization of a SMIng language, existing SMIv2 MIBs and SPPI PIBs on the standards track will not be required to be translated to the SMIng language. New MIBs/PIBs will be defined using the SMIng language. Motivation: Provide best-effort backwards compatibility for existing tools while not placing an unnecessary burden on MIBs/PIBs that are already on the standards track. 4.1.16 Base Data Types Number: 12 Type: basic From: SMI, SPPI Description: SMIng must support the base data types Integer32, Unsigned32, Integer64, Unsigned64, Enumeration, Bits, OctetString, and OID. Motivation: Most are already common. Unsigned64 and Integer64 are in SPPI, must fix in SMI. 4.1.17 Enumerations Number: 19 Type: basic From: SMI, SPPI Description: SMIng should provide support for enumerations. Enumerated values must be a part of the enumeration definition. Motivation: SMIv2 already has enumerated numbers and OIDs that can be used to identify things. Notes: Enumerations have the implicit constraint that the attribute is constrained to the values for the enumeration. 4.1.18 Discriminated Unions Number: 32 Type: should From: WG Description: SMIng must support a standard format for discriminated unions. Motivation: Allows to group related attributes together, such as InetAddressType (discriminator) and InetAddress, InetAddressIPv4, InetAddressIPv6 (union). The lack of discriminated unions has also lead to relatively complex sparse table work-around in some DISMAN mid-level manager MIBs. Notes: SMIng must support discriminated unions, and should support unions in general (Section 4.2.2). Discriminated unions have the implicit constraint that the union attribute type is constrained by the discriminator attribute. Elliott, et. al. Expires December 11, 2001 [Page 11] Internet-Draft SMIng Requirements June 2001 4.1.19 Instance Pointers Number: 14 Type: basic From: SPPI Description: SMIng must allow specifying pointers to instances (i.e., a pointer to a particular attribute in a row). Motivation: It is common practice in MIBs and PIBs to point to other instances. 4.1.20 Row Pointers Number: 15 Type: align From: SMI, SPPI Description: SMIng must allow specifying pointers to rows. A row pointer is a special case of an instance pointer. Motivation: It is common practice in MIBs and PIBs to point to other rows (see RowPointer, PIB-REFERENCES). 4.1.21 Constraints on Pointers Number: 23 Type: basic From: SPPI Description: SMIng must allow specifying the types of objects to which a pointer may point. Motivation: Allows code generators to detect and reject illegal pointers automatically. Can also be used to automatically generate more reasonable implementation-specific data structures. Notes: Pointer constraints are a special case of attribute value constraints (Section 4.3.3) in which the prefix of the OID (row or instance pointer) value is limited to be only from a particular table. 4.1.22 Base Type Set Number: 16 Type: basic From: SMI, SPPI Description: SMIng must support a fixed set of base types of fixed size and precision. The list of base types should not be extensible unless the SMI itself changes. Motivation: Interoperability. Elliott, et. al. Expires December 11, 2001 [Page 12] Internet-Draft SMIng Requirements June 2001 4.1.23 Extended Data Types Number: 13 & 18 combined Type: align From: SMI, SPPI Description: SMIng must allow a mechanism to allow types to be defined as new types that provide additional semantics (e.g., Counters, Gauges, Strings, etc.). It may be desirable to also allow the derivation of new types from derived types. New types must be as restrictive or more restrictive than the types that they are specializing. Motivation: SMI uses application types and textual conventions. SPPI uses derived types. 4.1.24 Units and Default Values of Defined Types Number: 65 Type: new From: NMRG Description: In SMIv2 OBJECT-TYPE definitions may contain UNITS and DEFVAL clauses and TEXTUAL-CONVENTIONs may contain DISPLAY-HINTs. In a similar fashion units and default values should be applicable to defined types and format information should be applicable to attributes. Motivation: Some MIBs introduce TCs such as KBytes and every usage of the TC then specifies the UNITS "KBytes". It would simplify things if the UNITS were attached to the type definition itself. Note that SMIng must clarify the behavior, if an attribute uses a defined type and both, the attribute and the defined type, have units/default/format information. 4.1.25 Arrays Number: 39 Type: new From: WG Description: SMIng should allow the definition of a SEQUENCE OF attributes or structures (Section 4.1.29). Motivation: The desire for the ability to have variable-length, multi-valued objects. Notes: There are fixed- and variable-size arrays, however fixed-size arrays are really just a constrained kind of variable-size arrays. Variable arrays can map to the EXPANDS clause - using the index of the parent and the index of the contained array table with the lifetime of the child table controlled by the parent table. The EXPANDS clause formally states that there is a 1:n existence relationship between tables and for the n instances to exist in Elliott, et. al. Expires December 11, 2001 [Page 13] Internet-Draft SMIng Requirements June 2001 the child table, the corresponding instance must exist in the parent table. Conceptually, arrays map to variable-sized tables in tables. If arrays are to be supported then the general problem of variable- sized tables in tables should be solved (vs. solving a very specific problem such as a fixed-sized table in table). 4.1.26 Tables Number: 25 Type: basic From: SMI, SPPI Description: SMIng should provide a mechanism for grouping attributes as tables. Motivation: 4.1.27 Table Existence Relationships Number: 26 Type: basic From: SMI, SPPI Description: SMIng should support INDEX, AUGMENTS, and EXTENDS. Motivation: These three table existence relationships that exist either in the SMIv2 or the SPPI. 4.1.28 Table Relationships Number: 48 Type: new From: WG Description: SMIng should support REORDERS clauses. Motivation: A REORDERS statement allows to swap indexing orders without having to redefine the whole table. Notes: The EXPANDS clause portion was removed from this requirement and is mentioned in Arrays (Section 4.1.25). 4.1.29 Structures Number: 33 Type: new From: NMRG Description: A structure is a non-divisible, extensible grouping of attributes that are meaningful together. Motivation: Required to map the same grouping of attributes into SNMP and COPS-PR tables. Allows to do index reordering without having to redefine the grouping of attributes. Allows to group related attributes together (e.g. InetAddressType, InetAddress). Elliott, et. al. Expires December 11, 2001 [Page 14] The ability to group attributes in a structure provides an indication that the attributes are meaningful together. Notes: Protocols must ensure that when sending structures across the wire that (1) all attributes of the structure are sent together and (2) the order of the attributes is maintained. 4.1.30 Containment Number: 40 Type: new From: NMRG Description: SMIng must provide support for the creation of new compound types from more basic (potentially compound) types. Motivation: Simplifies the reuse attribute combination such as InetAddressType and InetAddress pairs. Containment has the implicit existence constraint that if an instance of contained structure exists, then the corresponding instance of the containing structure must also exist. 4.1.31 Single Inheritance Number: 34 Type: new From: NMRG Description: SMIng should provide support for mechanisms to extend attribute groupings (structures) through single inheritance. Motivation: Allows to extend grouping of attributes, like a generic DiffServ scheduler, with attributes for a specific scheduler, without cut&paste. Notes: If an instance of a derived structure exists, then the corresponding instance of the base structure implicitly exists within the derived structure. Single inheritance with multiple levels (e.g., C derives from B, and B derives from A) must be allowed. Inheritance has the implicit existence constraint that if an instance of derived structure exists, then the corresponding instance of the base structure must also exist. 4.1.32 Events Number: 20 Type: basic From: SMI, SPPI Description: SMIng must provide mechanisms to define events which identify significant state changes. Motivation: These represent the protocol-independent events that lead to SMI notifications or SPPI reports. Elliott, et. al. Expires December 11, 2001 [Page 15] Internet-Draft SMIng Requirements June 2001 4.1.33 Creation/Deletion Number: 21 Type: basic From: SMI, SPPI Description: SMIng should support a mechanism to define creation/deletion operations for instances. Specific creation/deletion errors, such as INSTALL-ERRORS, must be supported. Motivation: Available for row creation in SMI, and available in SPPI. 4.1.34 Range and Size Constraints Number: 22 Type: basic From: SMI, SPPI Description: SMIng must allow specifying range and size constraints where applicable. Motivation: The SMI and SPPI both support range and size constraints. 4.1.35 Uniqueness Number: 24 Type: basic From: SPPI Description: SMIng must allow the specification of uniqueness constraints on attributes. SMIng should allow the specification of independent uniqueness constraints. Motivation: Knowledge of the uniqueness constraints on attributes allows to verify protocol specific mappings (e.g. INDEX clauses). The knowledge can also be used by code generators to improve generated implementation-specific data structures. 4.1.36 Extension Rules Number: 27 Type: basic From: SMI Description: SMIng must provide clear rules how one can extend SMIng modules without causing interoperability problems "over the wire". Motivation: SMIv2 and SPPI have extension rules. 4.1.37 Deprecate Use of IMPLIED Keyword Number: 30 Type: should From: SMI Description: The SMIng SNMP mapping should deprecate the use of the Elliott, et. al. Expires December 11, 2001 [Page 16] Internet-Draft SMIng Requirements June 2001 IMPLIED indexing schema. Motivation: IMPLIED is confusing and most people don't understand it. The solution (IMPLIED) is worse than the problem it is trying to solve and therefore for the sake of simplicity, the use of IMPLIED should be deprecated. 4.1.38 No Redundancy Number: 31 Type: should From: SMI Description: The SMIng language should avoid redundancy. Motivation: Remove any textual redundancy for things like table entries and SEQUENCE definitions, which only increase specifications without providing any value. 4.1.39 Compliance and Conformance Number: 50 Type: basic From: SMIv2, SPPI Description: SMIng should provide a mechanism for compliance and conformance specifications for protocol-independent definitions as well as for protocol mapping. Motivation: This capability exists in SMIv2 and SPPI. The NMRG proposal has the ability to express much of this information at the protocol-independent layer, thus reducing redundant information. Some compliance or conformance information may be protocol-specific, therefore there is also a need to be able to express this information in the mapping. 4.1.40 Allow Refinement of All Definitions in Conformance Statements Number: 74 Type: must From: Individual Description: SMIv2, RFC 2580, Section 3.1 says: The OBJECTS clause, which must be present, is used to specify each object contained in the conformance group. Each of the specified objects must be defined in the same information module as the OBJECT-GROUP macro appears, and must have a MAX-ACCESS clause value of "accessible-for-notify", "read-only", "read-write", or "read-create". The last sentence forbids to put a not-accessible INDEX object Elliott, et. al. Expires December 11, 2001 [Page 17] Internet-Draft SMIng Requirements June 2001 into an OBJECT-GROUP. Hence, you can not refine its syntax in a compliance definition. For more details, see http://www.ibr.cs.tu-bs.de/ietf/smi-errata/ Motivation: This error should not be repeated in SMIng. 4.2 Nice-to-Have Requirements This section represents the list of recommended requirements that would be nice to have. However, these are not automatically thought of as accepted requirements as, for example, they may entail a non- trivial amount of work in underlying protocols to support. 4.2.1 Methods Number: 37 Type: new From: WG Description: SMIng should support a mechanism to define method signatures (parameters, return values, exception) that are implemented on agents. Motivation: Methods are needed to support the definition of operational interfaces such as found in [RFC2925] (ping, traceroute and lookup operations). Also, the ability to define constructor/destructor interfaces could address issues such as encountered with SNMP's RowStatus solution. Notes: Is it possible to do methods without changing the underlying protocol? There is agreement that methods are useful, but disagreement upon the impact - one end of the spectrum sees this as a documentation tool for existing SNMP capabilities, while the other end sees this as a protocol update, moving forward, to natively support methods. The proposal is to wait and see if this is practical to implement as a syntax that is useful and can map to the protocol. 4.2.2 Unions Number: 32 Type: should From: WG Description: SMIng should support a standard format for unions. Motivation: Allows an attribute to contain one of many types of values. Allows related attributes to be grouped together. The lack of unions has also lead to relatively complex sparse table work-around in some DISMAN mid-level managers. Notes: The thought is that SNMP and COPS-PR can already support unions because they do not care about what data type goes with a particular OID. Elliott, et. al. Expires December 11, 2001 [Page 18] Internet-Draft SMIng Requirements June 2001 4.2.3 Abstract vs. Concrete Structures Number: 35 Type: new From: NMRG, WG Description: SMIng should differentiate between abstract and concrete grouping of attributes (structures). Motivation: This information gives people more information how structures can and should be used. It hinders them from misusing abstract structures. Notes: There is general confusion regarding the usefulness of abstract from the data point of view. This requirement attempted to convey the idea that some structures are not meant to stand on their own and instead only make sense if contained within another structure. The term abstract, which itself carries some connotation from the object-oriented world, may not be the best term to use. 4.2.4 Float Data Types Number: 49 Type: new From: WG, NMRG Description: SMIng should support the base data types Float32, Float64, Float128. Motivation: Missing base types can hurt later on, because they cannot be added without changing the language, even as an SMIng extension. Lesson learnt from the SMIv1/v2 debate about Counter64/Integer64/... Notes: There is no mention as to whether or not the underlying protocols will have to natively support float data types. This is left to the mapping. However, it seems imperative that the float data type needs to be added to the set of intrinsic types in the SMIng language at the creation of the language as it will be impossible to add them later without changing the language. 4.2.5 Comments Number: 59 Type: should From: NMRG Description: The syntax of comments should be well defined, unambiguous and intuitive to most people, e.g., the C++/Java `//' syntax. Motivation: ASN.1 Comments (and thus SMI and SPPI comments) have been a constant source of confusion. People use arbitrary lengthy strings of dashes (`-----------') in the wrong assumption that Elliott, et. al. Expires December 11, 2001 [Page 19] Internet-Draft SMIng Requirements June 2001 this is always treated as a comment. Some implementations try to accept these syntactically wrong constructs which even raises confusion. We should get rid of this problem. Notes: If the SMIng working group adopts a C-like language, then the C++/Java single-line comment should be adopted as well. 4.3 Rejected Requirements This section represents the list of requirements that were rejected because they either were deemed unnecessary, too complicated, outside the scope of the WG charter, or no-brainers. 4.3.1 Incomplete Translations Number: 10 Type: basic From: WG Description: Reality sucks. All information expressed in SMIng may not be directly translatable to a MIB or PIB construct, but all information should be able to be conveyed in documentation or via other mechanisms. Motivation: SMIng working group requires this to ease transition. Notes: Deemed not a valuable use of the working group's time. 4.3.2 Instance Naming Number: 11 Type: align From: SMI, SPPI Description: Instance naming is subject of the protocol mappings and not part of the protocol-neutral model. INDEX, PIB-INDEX must be accommodated. Motivation: COPS-PR and SNMP have different instance identification schemes that must be aligned in the protocol specific mappings. Notes: This requirement is not being rejected because it is a bad idea. It is rejected because MIBs and PIBs are sufficiently close to allow one specification, in the protocol-neutral model, for both. 4.3.3 Attribute Value Constraints Number: 44 Type: new From: WG Description: SMIng must provide mechanisms to formally specify constraints between values of multiple attributes. Motivation: Constraints on attribute values [occur] where one or more attributes may affect the value or range of values for another Elliott, et. al. Expires December 11, 2001 [Page 20] Internet-Draft SMIng Requirements June 2001 attribute. One such relationship exists in IPsec, where the type of security algorithm determines the range of possible values for other attributes such as the corresponding key size." Notes: This requirement as is has been rejected as too general, and therefore virtually impossible to implement. However, constraints that are implicit with discriminated unions (Section 4.1.18), enumerated types (Section 4.1.17), pointer constraints (Section 4.1.21)), etc., are accepted and these implicit constraints are mentioned in the respective requirements. 4.3.4 Existence Constraints Number: 41 Type: new From: WG Description: SMIng must provide a mechanism to formally express existence constraints. Motivation: Existence constraints are already embedded in SMIv2 INDEX clauses and DESCRIPTION clauses. Notes: The conclusion is that it is better to avoid explicit fate sharing and cover this with a description clause. Individual requirements (e.g., inheritance (Section 4.1.31), containment (Section 4.1.30), arrays (Section 4.1.25), etc.) that implicitly provide existence constraints have stated so in their requirements. 4.3.5 Ordering Constraints Number: 42 Type: new From: WG Description: SMIng must provide a mechanism to formally express ordering constraints. Motivation: Notes: It is not clear why this cannot be done in the description clause. 4.3.6 Attribute Transaction Constraints Number: 43 Type: new From: WG Description: SMIng must provide a mechanism to formally express that certain sets of attributes can only be modified in combination. Motivation: COPS-PR always does operations on table rows in a single transaction. There are SMIv2 attribute combinations that need to be modified together (such as InetAddressType, InetAddress). Notes: Alternative is to either use Methods (Section 4.2.1) or assume Elliott, et. al. Expires December 11, 2001 [Page 21] Internet-Draft SMIng Requirements June 2001 that all attributes in a structure (Section 4.1.29) are to be considered atomic. 4.3.7 Method Constraints Number: 47 Type: new From: WG Description: Method definitions must provide constraints on parameters. Motivation: Notes: Unless methods (Section 4.2.1) are done, there is no use for this. 4.3.8 Categories Number: 28 Type: basic From: SPPI Description: SMIng must provide mechanisms to group definitions into subject categories. Concrete instances may only exist in the scope of a given subject category or context. Motivation: To scope the categories to which a module applies. In SPPI this is used to allow a division of labor between multiple client types. Notes: This requirement is specific to COPS-PR and therefore does not have general applicability and may complicate matters as SNMP won't have much use for it. 4.3.9 Agent Capabilities Type: basic Number: 29 From: SMI Description: SMIng must provide mechanisms to describe agent implementations. Motivation: To permit manager to determine variations from the standard for an implementation. Notes: Agent capabilities should not be part of SMIng, but should instead be a separate capabilities table. 4.3.10 Relationships Number: 36 Type: new From: NMRG, WG Description: Ability to formally depict existence dependency, value dependency, aggregation, containment, and other relationships Elliott, et. al. Expires December 11, 2001 [Page 22] Internet-Draft SMIng Requirements June 2001 between attributes or groupings of attributes. Motivation: Helps humans to understand the conceptual model of a module. Helps implementers of MIB compilers to generate more `intelligent' code. Notes: This requirement was deemed too general to be useful and instead the individual types of relationship requirements (e.g., pointers, inheritance, containment, etc.) are evaluated on a case- by-case basis with the specific relationships deemed useful being included as accepted requirements. 4.3.11 Procedures Number: 38 Type: new From: WG Description: SMIng should support a mechanism to formally define procedures that are used by managers when interacting with an agent. Motivation: Notes: Best to do this in the description clause. 4.3.12 Associations Number: 45 Type: new From: WG Description: SMIng should provide mechanisms to explicitly specify associations. Motivation: 4.3.13 Association Cardinalities Number: 46 Type: new From: WG Description: Cardinalities between associations should be formally defined. Motivation: If you have an association between structures A and B, the cardinality of A indicates how many instances of A may be associated with a single instance of B. Our discussions in Minneapolis indicated that we want to convey "how many" instances are associated in order to define the best mapping algorithm - whether a new table, a single pointer, etc. For example, do we use RowPointer or an integer index into another table? Do we map to a table that holds instances of the association/relationship itself? Notes: Without associations (Section 4.3.12), this has no use. Elliott, et. al. Expires December 11, 2001 [Page 23] Internet-Draft SMIng Requirements June 2001 4.3.14 Categories of Modules Number: 51 Type: new From: Individual Description: The SMIng documents should give clear guidance on which kind of information (with respect to generality, type/structure/extension/..) should be put in which kind of a module. E.g., in SMIv2 we don't like to import Utf8String from SYSAPPL- MIB, but we also do not like to introduce a redundant definition. A module review process should probably be described that ensures that generally useful definitions do not go into device or service specific modules. Motivation: Bad experience with SMIv2. Notes: It is not clear how this can be done with the language to be created by SMIng WG. It could be analogous with header files, however there is potentially lots of process associated with doing this. There may be a better way to create TCs that are used by many. 4.3.15 Length of Identifiers Number: 53 Type: should From: NMRG Description: The allowed length of the various kinds of identifiers should be extended from the current `should not exceed 32' (maybe even from the `must not exceed 64') rule. Motivation: Reflect current practice of definitions. Notes: Not clear what value this provides, so decision was to keep things as is. 4.3.16 Why Are So Many SMIv1/v2 Parsers So Error Tolerant? Number: 54 Type: should From: Individual Description: It should be clearly stated that parser implementations which accept input that does not conform to the SMIng language rules are not compliant. Motivation: SMIv1/v2 parsers are tolerant, because MIB editors do not get SMI right, because it builds on hardly available obsolete ASN.1 CCITT specifications. With SMIng there is a chance to get the syntax clearly and self-contained defined, so that there is no Elliott, et. al. Expires December 11, 2001 [Page 24] Internet-Draft SMIng Requirements June 2001 excuse for errors and parser implementations become more consistent. (Who would expect that a C compiler is tolerant about a missing semicolon?) Notes: This requirement appears to be in conflict with language extensibility (Section 4.1.6). 4.3.17 Core Language Keywords vs. Defined Identifiers Number: 55 Type: should From: NMRG Description: In SMIv1/v2 things like macros or some(!) types have to be imported from SMI modules. People are continuously confused about what has to be imported (imagine if `typedef' would have to be #included in a C program) and what the difference between those SMI modules and usual modules is. Motivation: Reduce confusion. Clarify the set of language keywords. Notes: This requirement was not rejected because it was a bad idea. Instead, it was rejected as it was thought that it is basic enough of an idea that it's a no-brainer. The assumption can be made that any keywords defined in the SMIng language will not have to be imported. 4.3.18 Internationalization Number: 56 Type: new From: Individual Description: Informational text (DESCRIPTION, REFERENCE, ...) should allow i18nized encoding (UTF8? others?). Motivation: There has been some demand for i18n in the past. Notes: English is the language of the IETF and therefore it is not a requirement of the SMIng language must allow i18n. 4.3.19 Mapping Modules to Files Number: 58 Type: new From: NMRG Description: There should be a clear statement how SMIng modules are mapped to files (1:1, n:1?) and how files should be named (by module name in case of 1:1 mapping?). Motivation: SMI implementations show up a variety of filename extensions (.txt, .smi, .my, none). Some expect all modules in a single file, others don't. This makes it more difficult to exchange modules. Notes: This is just an implementation detail and is best left to a BCP and not made a part of the language definition. Elliott, et. al. Expires December 11, 2001 [Page 25] Internet-Draft SMIng Requirements June 2001 4.3.20 Simple Grammar Number: 60 Type: new From: NMRG Description: The grammar of the language should be as simple as possible. It should be free of exception rules. A measurement of simplicity is shortness of the ABNF grammar. Motivation: Ease of implementation. Ease of learning/understanding. Notes: This seems like an obvious requirement, however shortness of the ABNF grammar is not necessarily a reflection of the simplicity of the language. The WG will do the right thing with regard to defining the language for SMIng. 4.3.21 Place of Module Information Number: 61 Type: should From: NMRG Description: Module specific information (organization, contact, description, revision information) should be bound to the module itself and not to an artificial node (like SMIv2 MODULE-IDENTITY). Motivation: Simplicity and design cleanup. Notes: This does not seem to be a problem with the current SMI and is dropped for simplicity. 4.3.22 Fully Qualified Identifiers Number: 62 Type: should From: NMRG Description: To reference multiple identifiers with the same name but imported from multiple modules a qualifying mechanism, e.g., `module::name', is needed. It should be manifested in the grammar. (SMI and SPPI do support it already because of their ASN.1 derivation, but many implementation fail to handle this correctly.) Motivation: Unambiguous references to identifiers. Notes: Look at existing requirements for SMI and SPPI. Adding another requirement to SMIng for something that is already a requirement adds no value. 4.3.23 Readable Revision Date Representation Number: 63 Type: should From: NMRG Description: The SMI notation of revision dates consists of 11 or 13 Elliott, et. al. Expires December 11, 2001 [Page 26] Internet-Draft SMIng Requirements June 2001 characters, e.g. 199602282155Z, which is difficult to parse for humans. The trailing `Z' which represents GMT is superfluous, since no other time zone is allowed. SMIng should support a nicer notation, e.g. based on ISO 8601 representation: 1996-02-28 21:55 or simply 1996-02-28 since time information is not relevant in almost any case. Motivation: SMIng should support a nicer notation, Human readability of date and time information. Notes: There is no reason to make more complicated rules unless the syntax of the date string is to be changed completely. 4.3.24 Make Status Information Optional Number: 64 Type: new From: NMRG Description: SMI and SPPI definitions must have a status information (current, obsolete, deprecated). SMIng should make the status clause optional with a default of `current'. Furthermore, clear statements are required on constraints of status information of related definitions, e.g., a current attribute definition must not make use of an obsolete defined type, etc. Note, that this is problematic with definitions from multiple independently evolving modules. Motivation: Make definitions more compact. Hide redundant information. Notes This represents an insignificant change that just adds one more rule to follow. 4.3.25 Remove OIDs from the Core Language Number: 67 Type: new From: NMRG Description: While in SMI and SPPI definitions of attributes are bound to OIDs, SMIng should not use OIDs for the definition of structures, structure attributes, events, etc. Instead, SNMP and COPS-PR mappings should assign OIDs to the mapped items. Motivation: OIDs of synonymous attributes are not the same in SMI and SPPI definitions. Hence, they must not appear in protocol neutral definitions. Notes: Since both COPS-PR and SNMP both use OIDs, why not just align them. Clearly, structures and TCs should not contain OIDs, which is already true. The only problem with OIDs is that a structure (Section 4.1.29) cannot be reused if an OID is assigned at that level. The solution is the reusable components (e.g., structures) Elliott, et. al. Expires December 11, 2001 [Page 27] Internet-Draft SMIng Requirements June 2001 need not have OID numbers, while non-reusable components (e.g., tables) can have OIDs. 4.3.26 Module Namespace Number: 68 Type: new From: WG Description: Currently the namespace of modules is flat and there is no structure in module naming causing the potential risk of name clashes. Possible solutions: * Assume module names are globally unique (just as SMIv1/v2), just give some recommendations on module names. * Force all organizations, WGs and vendors to apply a name prefix (e.g. CISCO-GAGA-MIB, IETF-DISMAN-SCRIPT-MIB?). * Force enterprises to apply a prefix based on the enterprise number (e.g. ENT2021-SOME-MIB). * Put module names in a hierarchical domain based namespace (e.g. DISMAN-SCRIPT-MIB.ietf.org). Motivation: Reduce risk of module name clashes. Notes: Some aspects of this requirement overlapped with other requirements and other aspects were thought best left to a BCP. 4.3.27 Hyphens in Identifiers Number: 72 Type: should From: NMRG Description: There has been some confusion whether hyphens are allowed in SMIv2 identifiers: Module names are allowed to contain hyphens. Node identifiers usually are not. But for example `mib- 2' is a frequently used identifier that contains a hyphen due to its SMIv1 origin, when hyphen were not disallowed. Similarly, a number of named numbers of enumeration types contain hyphens violating an SMIv2 rule. SMIng should simply allow hyphens in all kinds of identifiers. No exceptions. Motivation: Reduce confusion and exceptions. Requires, however, that implementation mappings properly quote hyphens where appropriate. Notes: Since nobody is currently complaining about the hyphen problem, there is no reason to fix it. The restriction on "_" (underscore) should be relaxed. 4.3.28 Referencing a Group of Instances of a Structure Number: 75 Type: align Elliott, et. al. Expires December 11, 2001 [Page 28] Internet-Draft SMIng Requirements June 2001 From: Individual Description: PIB and MIB row attributes reference a group of entries in another table. This semantic needs to be formalized. Motivation: SPPI formalizes this feature using TagId and TagReferenceId semantics in the DiffServ PIB. SMI also uses these semantics without any formal notation (see SNMP-TARGET-MIB in RFC2273). Notes: There are no issues with the requirement, but instead with how this is mapped cleanly and more generally. Effectively, this is a constraint clause that specifies the table and column that contains a value used for grouping. This can be mapped to EXPANDS, which would have to be added to SPPI. 5. Security Considerations This document defines requirements for a language with which to write and read descriptions of management information. The language itself has no security impact on the Internet. 6. Acknowledgements Thanks to Dave Durham, whose work on the original NIM (Network Information Model) draft was used in generating this document. Thanks to Frank Strauss for maintaining the web site that was home for the requirements list during the requirements discussion period. Thanks to Andrea Westerinen for her contributions on the original NIM and SMIng drafts. References [1] Case, J., Fedor, M., Schoffstall, M. and J. Davin, "Simple Network Management Protocol (SNMP)", STD 15, RFC 1157, May 1990. [2] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [3] McCloghrie, K., Perkins, D., Schoenwaelder, J., Case, J., McCloghrie, K., Rose, M. and S. Waldbusser, "Structure of Management Information Version 2 (SMIv2)", STD 58, RFC 2578, April 1999. [4] McCloghrie, K., Perkins, D., Schoenwaelder, J., Case, J., McCloghrie, K., Rose, M. and S. Waldbusser, "Textual Conventions for SMIv2", STD 58, RFC 2579, April 1999. Elliott, et. al. Expires December 11, 2001 [Page 29] Internet-Draft SMIng Requirements June 2001 [5] McCloghrie, K., Perkins, D. and J. Schoenwaelder, "Conformance Statements for SMIv2", STD 58, RFC 2580, April 1999. [6] McCloghrie, K., Case, J., Rose, M. and S. Waldbusser, "Management Information Base for Version 2 of the Simple Network Management Protocol (SNMPv2)", RFC 1907, January 1996. [7] Wahl, M., Coulbeck, A., Howes, T. and S. Kille, "Lightweight Directory Access Protocol (v3): Attribute Syntax Definitions", RFC 2252, December 1997. [8] White, K., "Definitions of Managed Objects for Remote Ping, Traceroute, and Lookup Operations", RFC 2925, September 2000. [9] Chan, K., Seligson, J., Durham, D., Gai, S., McCloghrie, K., Herzog, S., Reichmeyer, F., Yavatkar, R. and A. Smith, "COPS Usage for Policy Provisioning (COPS-PR)", RFC 3084, March 2001. [10] McCloghrie, K., Fine, M., Seligson, J., Chan, K., Hahn, S., Sahita, R., Smith, A. and F. Reichmeyer, "Structure of Policy Provisioning Information (SPPI)", draft-ietf-rap-sppi-06.txt, April 2001. [11] Fine, M., McCloghrie, K., Seligson, J., Chan, K., Hahn, S., Sahita, R., Smith, A. and F. Reichmeyer, "Framework Policy Information Base", draft-ietf-rap-frameworkpub-04.txt, April 2001. [12] Strauss, F., Schoenwaelder, J. and K. McCloghrie, "SMIng - Next Generation Structure of Management Information", draft-irtf- nmrg-sming-04.txt, November 2000. Authors' Addresses Chris Elliott Cisco Systems 7025 Kit Creek Road Research Triangle Park, NC 27709 USA EMail: chelliot@cisco.com Elliott, et. al. Expires December 11, 2001 [Page 30] Internet-Draft SMIng Requirements June 2001 David Harrington Enterasys Networks 35 Industrial Way P.O. Box 5005 Rochester, NH 03866-5005 USA EMail: dbh@enterasys.com Jamie Jason Intel Corporation MS JF3-206 2111 NE 25th Ave. Hillsboro, OR 97124 USA EMail: jamie.jason@intel.com Juergen Schoenwaelder TU Braunschweig Bueltenweg 74/75 38106 Braunschweig Germany EMail: schoenw@ibr.cs.tu-bs.de URI: http://www.ibr.cs.tu-bs.de/ Frank Strauss TU Braunschweig Bueltenweg 74/75 38106 Braunschweig Germany EMail: strauss@ibr.cs.tu-bs.de URI: http://www.ibr.cs.tu-bs.de/ Walter Weiss Ellacoya Networks 7 Henry Clay Dr. Merrimack, NH. 03054 USA EMail: wweiss@ellacoya.com Elliott, et. al. Expires December 11, 2001 [Page 31] Internet-Draft SMIng Requirements June 2001 Appendix A. Mailing List Discussions o Human Readability (Section 4.1.3): * Jamie: One thing I have noticed reading the IPsec PIB and MIB documents is that the semantics of the model being presented are easily lost. The reason this is important to me is as a co-author on the IPsec policy model, I want to make sure that the PIB and MIB are semantically equivalent to the policy model so that they represent the same information. Parsing through the SMI and SPPI to understand the semantics of the particular derivation for me is excruciatingly painful. I think that a language that is more akin to C would make that human parsing of the PIB/MIB much easier. As it stands now, I am relegated to drawing pictures of the tables in order to understand what is going on. * Todd A Anderson: I would highly prefer a C-like syntax (or a CORBA IDL-like syntax given than CORBA IDL is similar to C-like syntax) to an ANS1-like syntax. I think that C-like syntax is clearer and more straight-forward. * David Putzolu in consequence of his comments on #37 and #45-46: Finally, if I were brave enough to try to satisfy the above mentioned requirements in SMIng, doing so and expressing the relevant syntax in ASN.1 sounds extremely painful - if we must go down this path, lets do it using something C++, or Java-like - that would at least give a syntactic foundation that is relevant to writing a programming language. o Instance Pointers (Section 4.1.19): * Jamie: It is common when data modeling to reference another object instead of embedding the referenced object inside of the object doing the referencing. This is also important as it allows objects to have independent lifetimes as well as be referenced by many objects. * Jamie: Can someone please elaborate on the differences between #14 (Instance Pointers) and #15 (Row Pointers). Can they be thought of in this way? Instance pointers are typed pointers, whereas row pointers are void pointers. With the distinction being that an instance pointer may only reference a row in one type of table, while a row pointer may reference a row in any kind of table? * Juergen: I think the terms are used as defined in RFC 2579 (InstancePointer and RowPointer). The RowPointer always points to an instance of a row while the InstancePointer can also point to a particular cell in a table. Note that this has nothing to do with a typed pointer (a concept which does not really exist in SMIv2). SPPI however has ways to type pointers - that is to restrict them so that they can not point to everything. (Using a class-based terminology, the difference is whether we Elliott, et. al. Expires December 11, 2001 [Page 32] Internet-Draft SMIng Requirements June 2001 are done with supporting pointers to class instances or whether we also need pointers to concrete attributes of class instances.) o Accessibility (Section 4.1.5): * Frank: To some degree accessibility might depend on the protocol. Creation/deletion information might be protocol dependent(?). `Accessible for notify' might be protocol dependent(?). Do we need `write-only'? * Robert Story: There is a recent thread over in the SNMPv3 list in this vein. Someone asked what they should return for a set- able password object: asterisks, empty string, NULL, etc. A syntax of write-only would help in this case. * David H.: Row creation/deletion is a fundamental feature of SMIv2 and is required for backward compatibility and interoperability. Accessible-for-notify is a feature of SMIv2 and must be supported for backwards compatibility and interoperability. * David Perkins: Row creation/deletion is not really part of the SMI except for the status of read-create. Otherwise, the SMI is blissfully ignorant of creation and deletion. You may claim that the RowStatus TC makes creation and deletion part a fundamental feature. If so, I disagree, since the SMI does not require one to use the RowStatus TC for creation/deletion and when RowStatus is used, there is nothing special about the table and columns that use it. * David Perkins: On status accessible-for-notify, this is a status value whose usage and interpretation has been twisted since its creation and is abused in almost every case where it is used. It increases the difficulty of testing, and can easily result in errors that show up only during exceptional situations. * Frank: Let's be precise: Not creation/deletion is a feature of SMIv2, but a notation that allows to express whether a table allows creation/deletion of instances through protocol operations. I agree that both, create/delete information and accessible-for-notify information, is required in SMIng, but I'm not sure whether we need it in the protocol neutral or in the protocol dependant parts. o Creation/Deletion (Section 4.1.33): * Frank: Let's not mix up protocol operations and the data model. SMI does not have what this issue demands and IMHO this issue's status should be `new' instead of `basic'. * David H.: draft-ietf-rap-sppi-06.txt has Install-ERRORS as part of the SPPI grammar. Both SMI and SPPI discuss "read-create". rfc2079.txt discusses using the MAX-ACCESS clause to indicate whether it makes protocol sense to create an instance of an object. As I see it, both recognize the need to be able to create "things" in the protocol. RFC2079 has the RowStatus T-C Elliott, et. al. Expires December 11, 2001 [Page 33] Internet-Draft SMIng Requirements June 2001 which explicitly discusses how to delete a "thing" in the protocol. SPPI, in the description of the ReferenceID, explicitly discusses deleting an instance of a PRI. All of these discussions of creation/deletion are found in the SMI/SPPI documents, not in the SNMP or COPS-PR protocol documents. * Frank: Ok. So the first sentence of this requirement's description should be reworded to something like: SMIng should support a mechanism to define whether creation/deletion protocol operations for instances would make sense. This is already covered by issue #17. o Categories (Section 4.3.8) * Jamie: Is this akin to C++ namespaces (apologies to those who are not familiar with them) in that they allow for scoping in order to reduce/prevent name collisions? Or, is this the purpose of #68 (Module Namespace)? If #68 serves this purpose, I would like some more clarification on #28 so that I can get them straight in my mind. * Juergen: This #28 is not about C++ namespaces. Categories are a way to categorize definitions - e.g. all the definitions relevant for a diffserv manager or all the definitions relevant for the security manager. COPS-PR and SPPI have this concept. The SNMP world sometimes uses contexts to achieve something similar. I once had a long debate with Keith about the difference between contexts and subject categories and I am still confused about it. ;-) o Agent Capabilities (Section 4.3.9) * Frank: Capability statements in MIB modules are hardly useful to managers, since in most cases they are simply not available. Agent capabilities should be retrievable at runtime from the agent itself through something like a capabilities MIB. Hence, I suggest to drop `agent capabilities' from the core SMIng language. * Jamie: I would like to second the idea that #29 (Agent Capabilities) be removed from the SMIng requirements. This does not seem to belong at this level. o Deprecate Use of IMPLIED Keyword (Section 4.1.37): * Frank: The SNMP mapping must keep IMPLIED for compatibility. But it should be clearly stated that it must not be used in newly defined SNMP mappings. o Classes (Section 4.1.29): * Jamie: I agree with the motivation - I think it is a good thing to be able to group attributes together for reuse. However, I am wondering if the name classes can be changed to something more generic. I don't know if "structures" are any better, but I would like to see a different description. * Jamie later on: Would it be possible to name #33 (Classes) to something like "Attribute Groups"? Elliott, et. al. Expires December 11, 2001 [Page 34] Internet-Draft SMIng Requirements June 2001 * Juergen: Fine with me. * David Putzolu: Methods are strongly associated with classes in the programming language lexicon, so choosing a different name is a good idea. * David D.: How about Attribute Class or aclass for short? Given your comment to #34 below, I think the word "class" implies inheritance abilities, whereas Attribute Groups, Structures, etc. do not. o Single Inheritance (Section 4.1.31): * Frank: I strongly suggest to decide whether this is a requirement after a few examples are found including their mappings to SNMP and COPS-PR tables! If we look at the example above, each inherited scheduler would have to remain a scheduler whose instances all appear in a common scheduler table. Thus the mapping to SNMP should lead to a basic table that holds all common attributes (the parent class?) and a number of table augmentations. Currently, I see now way how inheritance could help in this situation. Cut&paste is not needed. Table relationships are the key. * Frank: Another motivation: inheritance could help to add attributes to a class that are specific to certain protocol mapping and do not appear in the protocol neutral module. E.g., RowStatus attributes in SNMP mappings. * Jamie: I see this important because as more WGs move to data modeling, it is natural to model using OO methodologies. For example, in the IPsec Policy WG we are modeling the IPsec configuration policy (draft-ietf-ipsp-config-policy-model- 02.txt), which derives from the Policy Core Information Model from the Policy Framework WG. Both are modeled using OO methodology and make extensive use of single inheritance. In addition to the abstract model, the WG is defining a PIB (draft-ietf-ipsp-ipsecpib-02.txt) and a MIB (draft-ietf-ipsp- ipsec-conf-mib-00.txt) as concrete instantiations of the abstract model. * David Putzolu: As long as we avoid methods, ctors/dtors, exceptions, etc., then single inheritance is not only useful but feasible as well. o Abstract vs. Concrete Classes (Section 4.2.3) * Jamie: When doing data modeling using OO methodologies, it is important to be able to define an abstract class, which contains some set of attributes common to all derived classes, but which is never meant to be instantiated by itself. Again, an example is the IPsec policy configuration model - in that model, we have the idea of an IPsec transform. There are current three transforms in the model - AH, ESP, and IPCOMP. All three share a set of attributes. Instead of repeating the definitions of these attributes in each derived class, the attributes are defined in an abstract base class and all three Elliott, et. al. Expires December 11, 2001 [Page 35] Internet-Draft SMIng Requirements June 2001 derive from the abstract base class. o Methods (Section 4.2.1) * David Putzolu: If I define a class (attribute grouping?) that derives from an abstract class, and the abstract class defines a method with a return value of one type, but my newly defined class has a method of the same name but returns another type, which one is invoked? Questions like this worry me - but I can probably answer this question if the language being used is C++ or Java. However, the goal is not to do all the things that C++ or Java does (come up with a programming language) - it is to make a modeling language. As such, on requirement #37 I strongly agree with Juergen when he wrote, "I prefer to stay away from methods at this point in time (but make the language extensible so they might be added later)." * Juergen: The WG charter says that we do a "next generation data definition language for specifying network management data". We are not supposed to create a generic (information) modeling language. My argument for postponing methods is based on the observation that SNMP as well as COPS currently do not support method invocation natively and so you either have a feature in the data definition language which you can't use in practice (at least with SNMP and COPS-PR) or you have to do really ugly things to emulate generic method calls on top of what SNMP and COPS provides you. The issue of naming scopes and how you resolve conflicts is important. But I do not think it is very complex to handle this nor do I think that specifying how you solve name conflicts makes the data definition language a programming language. We already have rules in the SMIv2/SPPI how to resolve conflicts if you have to import definitions with similar names. If we have methods, we sure need more rules - but nothing conceptually very different. * David Putzolu again on #37 revisited: If we were to do methods, that of course leads to the idea of exceptions. Exceptions are a great tool for writing high quality software for a number of reasons - but I have no clue how they relate to the case of a data modeling language that will be mapped to on-the-wire SNMP and COPS PDUs. How would I map a try..catch block or a throw() to a SNMP PDU? What meaning does the idea of passing an unhandled exception on to a higher execution context (stack unrolling) have to a DECision message? I strongly suggest that if methods are done that exceptions not be done. If methods are not done, the question of exceptions becomes moot. * Juergen: Exceptions model exceptional conditions that can happen on the "agent" while you invoke a "method". A good example are the INSTALL-ERRORS in the SPPI which enumerate the exceptions that can happen while creating a new row via COPS- Elliott, et. al. Expires December 11, 2001 [Page 36] Internet-Draft SMIng Requirements June 2001 PR. Note that this notion of exceptions is completely independent from the way an application written in a particular programming language processes exceptions. The try...catch block is a programming language construct and we are of course not going to specify this. If you want an example how this can work, then please take a look at the CORBA world. The CORBA folks define exceptions at the IDL level. The programming language binding later says how things are mapped to programming language constructs. And the mappings look very different, depending on whether you use C or Java. o Arrays (Section 4.1.25): * Juergen: It is unclear what this really means. Does an array imply atomic access to the whole array? Or is it sufficient to say this is just a short-cut for another expanding table? * Andrea: I had viewed this only as allowing a multi-valued attribute, not as arrays of multiple attributes. This needs to be clarified. * Todd A Anderson: I prefer the IDL verbiage of "sequence" instead of array since array implies something of fixed length to me. I think that sequences are a necessary part of SMIng since I am constantly frustrated by the proliferation of tables I have to create to simulate sequence semantics. It seems to me that without sequences, the number of associations would also have to increase. I think it is just clearer and more natural for those with a programming background to think of sequences of data instead of breaking up the data structure into multiple locations. * Jamie: Is this as Andrea thought (a multi-valued attribute), or is it a set of multiple attributes? I can see use for having a set of multiple attributes (which is what I had thought it was). * Juergen: My understanding is that this refers to a multi-valued attribute. o Ordering Constraints (Section 4.3.5) * Frank: What does this mean? * Andrea: At least in some of the discussions, this "requirement" went hand in hand with #43 (transaction constraints). IE, if you modify something "in combination", the changes may need to occur in a specific order (first attribute A, then attribute B). o Attribute Transaction Constraints (Section 4.3.6) * Todd A Anderson: Could someone provide some clarity on issues #42 and #43? An issue that seems similar to #43 is the case when sometimes I want to execute several table changes atomically but other times I may not want to make those changes atomically. Is the language an appropriate place to deal with Elliott, et. al. Expires December 11, 2001 [Page 37] Internet-Draft SMIng Requirements June 2001 this sort of behavior? Are there any modeling issues related to this sort of transaction? My instinct is that there isn't for this type of transaction but for the case where you must always change several things atomically then the language is an appropriate place to state that. How would the other type of transaction be handled then? o Associations (Section 4.3.12) * Frank: What does this mean? Aren't relationships (issue #36) the same? * Andrea: Yes, an association is a kind of relationship but has additional info like cardinality on the related entities. Relationships include inheritance. * David Putzolu on #45 and #46: These are two more ideas that on their own make good sense, but seem to complicate the big picture. How would cardinality be captured in a mapping to SNMP or COPS? Pointers seems pretty easy to map to these protocols - but where does associations fit in? These two are elegant tools, but I think in this context, since we already have pointers, and two pointers in a table can model an association, simplicity says remove these two requirements. o Association Cardinalities (Section 4.3.13) * See also David Putzolu's comment on issue #45. o Float Data Types (Section 4.2.4) * Todd A Anderson: I am glad to see that the spec includes float data types in the language. I find these types especially useful for TSPECs and fractional link bandwidth partitioning. o Why Are So Many SMIv1/v2 Parsers So Error Tolerant? (Section 4.3.16) * David H.: I think the reality of the situation is that developers write few mibs, but write much C code. There are few developers who understand mib syntax, and I don't expect to see that improve even if we use a non-ASN.1 language, and parsers are classified as non-compliant. I won't oppose this suggestion, but I doubt it will solve the problem. * Frank: If parsers are (available and) forced to be verbatim about errors then MIB authors have simple tools to validate their modules. C programs are correct because they must be compiled to be useful and because C compilers are strict. I agree, that many people are not really familiar with MIB syntax because they write much less MIB modules than C (or other) code. And I agree that a non-ASN.1 looking syntax would not help significantly. o Internationalization (Section 4.3.18) * David H.: Fred Baker made it very clear as IESG chair that all documents submitted for standards advancement should be done in English to ensure a large enough body of reviewers exists to provide industry-wide review. Elliott, et. al. Expires December 11, 2001 [Page 38] Internet-Draft SMIng Requirements June 2001 I fear making mib descriptions i18n capable would encourage development of mibs written in languages that most developers could not read, and that would hurt interoperability. I understand that it is frustrating for those who do not speak English as their primary language to be forced to use English. However, the purpose of standards is to improve interoperability. * Randy: Many MIBs are never subjected to the standards process. A specification should be intelligible to the community of developers and users that will use that MIB. Many MIBs never see use outside the organizations that defined them. Let the RFC submission / publication process do its job of weeding out horrible things like the name of the city where I live, and tricky things like non-English words and phrases that fit into seven-bit ASCII. We don't have to replicate that service in our language definition, just as our language definition doesn't need to recapitulate the I-D and RFC rules for page breaks. We are not doing the world a service by preventing organizations from using the tools we define to develop specifications that their developers can understand. o Mapping Modules to Files (Section 4.3.19) * David H.: I think this is two separate requirements with different potential effects on the community and should be described separately. I have no issue with deciding that there should only be one module per file, or that more than one can be bundled together. (I prefer the single module per file to make updates easier) I am concerned that requiring specific filename formatting may prevent files from being used on some operating systems. I gladly accept that the documents should recommend, but not require, a consistent format for naming mib files. But I would consider it a bad thing to have a compiler refuse to compile a mib because the filename doesn't match the mib name, or whatever. o Place of Module Information (Section 4.3.21) * David H.: I don't understand what is being requested here. Is the (like SMIv2 MODULE-IDENTITY) and example if what is desired, or an example what is not desired? Where does module information belong in the proposer's eyes? * Frank: I'm sorry for the confusion. I try to be more precise: In SMIv2 and SPPI the MIB/PIB author has to put some module meta information in a specific macro (MODULE-IDENTITY) which is Elliott, et. al. Expires December 11, 2001 [Page 39] Internet-Draft SMIng Requirements June 2001 registered with an OID although this registration is not useful for any purpose. The proposal is to let SMIng (a) not register module meta information in the registration tree and (b) not introduce a new macro/statement wrapping the module meta information, since the module itself is the appropriate container. o Make Status Information Optional (Section 4.3.24) * David H.: I believe this is a bad idea if we allow inheritance and independent evolution of modules. It is very possible that a base class could be declared obsolete, but the derived classes would still incorrectly default to current. Some might conclude that they cannot obsolete something that somebody might have derived from, so they leave it marked as current. I think defaulting to current will be very confusing to people and the cure is worse than the illness. We need to make our standards unambiguous, much more than we need to eliminate a little redundancy. * Frank: Status information w.r.t. inheritance is a general problem as already stated in the description section. It does not matter whether the status clause is optional with a well defined default status if absent or whether the status clause is mandatory. There is no problem of ambiguity. * Jamie: Am I right in assuming that #64 (Make Status Information Optional) refers to status information that is most useful to a human? For example, if something is deprecated, a compiler could inform the user that they are depending on/deriving from/referencing something that has been deprecated in a manner similar to how the Java compiler does. If we go down the path of keeping the status information for the purpose of providing meaningful information from compilers, do we go down the road of also supplying additional information. For example, in the case of a deprecated class that is inherited from, should there also be information that states the name of the new class that should be inherited from instead? * Juergen: Issue #64 only deals with the proposal to make the status statement in the language optional in order to make definitions more compact and easier to read for humans. Issue #64 does not propose to change the semantics of the status values as they are used in the SMIv2 or the SPPI. What compilers do with the status values is implementation specific. Sure, a good compiler should warn if current definitions depend on deprecated or obsolete definitions. I personally would leave it to the MIB authors who deprecates definitions to explain the situation in the description clause. I personally prefer to not add language complexity in this case as the benefit does not seem clear/convincing to me. Elliott, et. al. Expires December 11, 2001 [Page 40] Internet-Draft SMIng Requirements June 2001 o Allow Refinement of All Definitions with Conformance Statements (Section 4.1.40): * David H.: I am not aware that this has been a problem except for one person. I am concerned that the requested requirement be that "All Definitions" must be allowed to be refined rather than to request that the one specific problem be addressed. o Referencing a Group of Instances of a Class (Section 4.3.28) * David H.: Is this already covered by #46 Association Cardinalities? Do we need to separate the formal specification of cardinality from its use here for associations? Elliott, et. al. Expires December 11, 2001 [Page 41] Internet-Draft SMIng Requirements June 2001 Full Copyright Statement Copyright (C) The Internet Society (2001). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Acknowledgement Funding for the RFC Editor function is currently provided by the Internet Society. Elliott, et. al. Expires December 11, 2001 [Page 42]