Network Working Group
Request for Comments: 1076
Obsoletes: RFC 1023
G. Trewitt
Stanford University
C. Partridge
November 1988

HEMS Monitoring and Control Language


1.   Status of This Memo                                               1
     Introduction                                                      2
2.   Overview and Scope                                                2
3.   Overview of Query Processor Operation                             4
4.   Encoding of Queries and Responses                                 5
4.1  Notation Used in This Proposal                                    5
5.   Data Organization                                                 6
5.1  Example Data Tree                                                 7
5.2  Arrays                                                            8
6.   Components of a Query                                             9
7.   Reply to a Query                                                 10
8.   Query Language                                                   12
8.1  Moving Around in the Data Tree                                   14
8.2  Retrieving Data                                                  15
8.3  Data Attributes                                                  16
8.4  Examining Memory                                                 18
8.5  Control Operations:  Modifying the Data Tree                     19
8.6  Associative Data Access:  Filters                                21
8.7  Terminating a Query                                              26
9.   Extending the Set of Values                                      27
10.  Authorization                                                    27
11.  Errors                                                           28
I.   ASN.1 Descriptions of Query Language Components                  29
I.1  Operation Codes                                                  30
I.2  Error Returns                                                    31
I.3  Filters                                                          33
I.4  Attributes                                                       34
I.5  VendorSpecific                                                   36
II.  Implementation Hints                                             36
III. Obtaining a Copy of the ASN.1 Specification                      42


This RFC specifies a query language for monitoring and control of network entities. This RFC supercedes RFC-1023, extending the query language and providing more discussion of the underlying issues.

This language is a component of the High-Level Entity Monitoring System (HEMS) described in RFC-1021 and RFC-1022. Readers may wish to consult these RFCs when reading this memo. RFC-1024 contains detailed assignments of numbers and structures used in this system. Portions of RFC-1024 that define query language structures are superceded by definitions in this memo. This memo assumes a knowledge of the ISO data encoding standard, ASN.1.

Distribution of this memo is unlimited.


This RFC specifies the design of a general-purpose, yet efficient, monitoring and control language for managing network entities. The data in the entity is modeled as a hierarchy and specific items are named by giving the path from the root of the tree. Most items are read-only, but some can be "set" in order to perform control operations. Both requests and responses are represented using the ISO ASN.1 data encoding rules.


The basic model of monitoring and control used in this memo is that a query is sent to a monitored entity and the entity sends back a response. The term query is used in the database sense -- it may request information, modify data, or both. We will use gateway- oriented examples, but it should be understood that this query- response mechanism is applicable to any IP entity.

In particular, there is no notion of an interactive "conversation" as in SMTP [RFC-821] or FTP [RFC-959]. A query is a complete request that stands on its own and elicits a complete response.

In order to design the query language, we had to define a model for the data to be retrieved by the queries, which required some understanding of and assumptions to be made about the data. We ended up with a fairly flexible data model, which places few limits on the type or size of the data.

Wherever possible, we give motivations for the design decisions or assumptions that led to particular features or definitions. Some of the important global considerations and assumptions are:

  • The query processor should place as little computational burden on the monitored entity as possible.
  • It should not be necessary for a monitored entity to store the complete query. Nothing in the query language should preclude an implementation from being able to process the query on the fly, producing portions of the response while the query is still being read and parsed. There may be other constraints that require large amounts of data to be buffered, but the query language design must not be one.
  • It is assumed that there is some mechanism to transport a sequence of octets to a query processor within the monitored entity and that there is some mechanism to return a sequence of octets to the entity making the query. In HEMS, this is provided by HEMP and its underlying transport layer. The query language design is independent of these details, however, and could be grafted onto some other protocol.
  • The data model must provide organization for the data, so that it can be conveniently named.

- Much of the data to be monitored will be contained in

tables. Some tables may contain other tables. The query

language should be able to deal with such tables.

  • We don't provide capabilities for data reduction in the query language. We will provide for data selection, for example, only retrieving certain table entries, but we will not provide general facilities for processing data, such as computing averages.
  • Because one monitoring center may be querying many (possibly hetrogenous) hosts, it must be possible to write generic queries that can be sent to all hosts, and have the query elicit as much information as is available from each host. i.e., queries must not be aborted just because they requested non-existent data.

There were some assumptions that we specifically did not make:

  • It is up to the implementation to choose what degree of concurrency will be allowed when processing queries. By locking only portions of the database, it should be possible to achieve good concurrency while still preventing deadlock.
  • This specification makes no statement about the use of the "definite" and "indefinite" length forms in ASN.1. There is currently some debate about this usage in the ISO community; implementors should note the recommendations in the ASN.1 specification.

Other RFCs associated with HEMS are:

      RFC-1021        Overview;
      RFC-1022        Transport protocol and message encapsulation;
      RFC-1024        Precise data definitions.

The rest of this report is organized as follows:

      Section 3       Gives a brief overview of the data model and the
                      operation of the query processor.
      Section 4       Describes the encoding used for queries and
                      responses, and the notation used to represent them
                      in this report.
      Section 5       Describes how the data is organized in the
                      monitored entity, and the view provided of it by
                      the query processor.
      Section 6       Describes the basic data types that may be given
                      to the query processor as input.
      Section 7       Describes how a reply to a query is organized.
      Section 8       Describes the operations available in the query
      Section 9       Describes how the set of data in the tree may be
      Section 10      Describes how authorization issues affect the
                      execution of a query.
      Section 11      Describes how errors are reported, and their
                      effect on the processing of the query.
      Appendix I      Gives precise ASN.1 definitions of the data types
                      used by the query processor.
      Appendix II     Gives extensive implementation hints for the core
                      of the query processor.


In this section, we give an overview of the operation of the query processor, to provide a framework for the later sections.

The query language models the manageable data as a tree, with each branch representing a different aspect of the entity, such as different layers of protocols. Subtrees are further divided to provide additional structure to the data. The leaves of the tree contain the actual data.

Given this data representation, the task of the query processor is to traverse this tree and retrieve (or modify) data specified in a query. A query consists of instructions to move around in the tree and to retrieve (or modify) named data. The result of a query is an exact image of the parts of the tree that the query processor visited.

The query processor is very simple -- it only understands eight commands, most of which share the same structure. It is helpful to think of the query processor as an automaton that walks around in the tree, directed by commands in the query. As it moves around, it copies the tree structure it traverses to the query result. Data that is requested by the query is copied into the result as well. Data that is changed by a query is copied into the result after the modification is made.


Both queries and responses are encoded using the representation defined in ISO Standard ASN.1 (Abstract Syntax Notation 1). ASN.1 represents data as sequences of <tag,length,contents> triples that are encoded as a stream of octets. The data tuples may be recursively nested to represent structured data such as arrays or records. For a full description, see the ISO standards IS 8824 and IS 8825. See appendix for information about obtaining these documents.

4.1 Notation Used in This Proposal

The notation used in this memo is similar to that used in ASN.1, but less formal, smaller, and (hopefully) easier to read. We will refer to a <tag,length,contents> tuple as a "data object". In this RFC, we will not be concerned with the details of the object lengths. They exist in the actual ASN.1 encoding, but will be omitted in the examples here.

Data objects that have no internal ASN.1 structure such as integer or octet string are referred to as "simple types" or "simple objects". Objects which are constructed out of other ASN.1 data objects will be referred to as "composite types" or "composite objects".

   The notation
   represents a simple object whose tag is "ID" with the given value.  A
   composite object is represented as
       ID{ ... contents ... }
   where contents is a sequence of data objects.  The contents may
   include both simple and structured types, so the structure is fully
   The difference between simple and composite types is close to the
   meaning of the "constructor" bit in ASN.1.  For the uses here, the
   distinction is made based upon the semantics of the data, not the
   representation.  Therefore, even though an OctetString can be
   represented in ASN.1 using either constructed or non-constructed
   forms, it is conceptually a simple type, with no internal structure,
   and will always be written as
       ID("some arbitrary string")
   in this RFC.
   There are situations where it is necessary to specify a type but give
   no value, such as when referring to the name of the data.  In this
   situation, the same notation is used, but with the value omitted:
       ID   or  ID()   or   ID{}
   Such objects have zero length and no contents.  The latter two forms
   are used when a distinction is being made between simple and
   composite data, but the difference is just notation -- the
   representation is the same.

ASN.1 distinguishes between four "classes" of tags: universal, application-specific, context-dependent, and reserved. HEMS and this query language use the first three. Universal tags are assigned in the ASN.1 standard and its addendums for common types, and are understood by any application using ASN.1. Application-specific tags are limited in scope to a particular application. These are used for "well-known" identifiers that must be recognizable in any context, such as derived data types. Finally, context-dependent tags are used for objects whose meaning is dependent upon where they are encountered. Most tags that identify data are context-dependent.


Data in a monitored entity is modeled as a hierarchy. Implementations are not required to organize the data internally as a hierarchy, but they must provide this view of the data through the query language. A hierarchy offers useful structure for the following operations:

   Organization    A hierarchy allows related data to be grouped
                   together in a natural way.
   Naming          The name of a piece of data is just the path from the
                   root to the data of interest.

Mapping onto ASN.1

ASN.1 can easily represent a hierarchy by using a "constructor" type as an envelope for an entire subtree.

Efficient Representation

Hierarchical structures are compact and can be traversed quickly.

   Safe Locking    If it is necessary to lock part of the hierarchy (for
                   example, when doing an update), locking an entire
                   subtree can be done efficiently and safely, with no
                   danger of deadlock.

We will use the term "data tree" to refer to this entire structure. Note that this internal model is completely independent of the external ASN.1 representation -- any other suitable representation would do. For the sake of efficiency, we do make a one-to-one mapping between ASN.1 tags and the (internal) names of the nodes. The same could be done for any other external representation.

Each node in the hierarchy must have names for its component parts. Although we would normally think of names as being ASCII strings such as "input errors", the actual name is just an ASN.1 tag. Such names are small integers (typically, less than 30) and so can easily be mapped by the monitored entity onto its internal representation.

We use the term "dictionary" to mean an internal node in the hierarchy. Leaf nodes contain the actual data. A dictionary may contain both leaf nodes and other dictionaries.

5.1 Example Data Tree

Here is a possible organization of the hierarchy in an entity that has several network interfaces and does IP routing. The exact organization of data in entities is specified in RFC-1024. This skeletal data tree will be used throughout this RFC in query examples.

System {

                  name                            -- host name
                  clock-msec                      -- msec since boot
                  interfaces                      -- # of interfaces
          Interfaces {                            -- one per interface
                  InterfaceData{ address, mtu, netMask, ARP{...}, ... }
                  InterfaceData{ address, mtu, netMask, ARP{...}, ... }
          IPRouting {
                  Entry{ ip-addr, interface, cost, ... }
                  Entry{ ip-addr, interface, cost, ... }

There are three top-level dictionaries in this hierarchy (System, Interfaces, and IPRouting) and three other dictionary types (InterfaceData, Entry, and ARP), each with multiple instances.

      The "name" of the clock in this entity would be:
          system{ clock-msec }
      and the name of a routing table entry's IP address would be:
          IPRouting{ Entry{ ip-addr } }.
      More than one piece of data can be named by a single ASN.1 object.
      The entire collection of system information is named by:
      and the name of a routing table's IP address and cost would be:
          IPRouting{ Entry{ ip-addr, cost } }.

5.2 Arrays

There is one sub-type of a dictionary that is used as the basis for tables of objects with identical types. We call these dictionaries arrays. In the example above, the dictionaries for interfaces, routing tables, and ARP tables are all arrays.

In the examples above, the "ip-addr" and "cost" fields are named. In fact, these names refer to the field values for ALL of the routing table entries -- the name doesn't (and can't) specify which routing table entry is intended. This ambiguity is a problem wherever data is organized in tables. If there was a meaningful index for such tables (e.g., "routing table entry #1"), there would be no problem. Unfortunately, there usually isn't such an index. The solution to this problem requires that the data be accessed on the basis of some of its content. Filters, discussed in section 8.6, provide this mechanism.

The primary difference between arrays and plain dictionaries is that arrays may contain only one type of item, while dictionaries, in general, will contain many different types of items. For example, the dictionary IPRouting (which is an array) will contain only items of type Entry.

The fact that these objects are viewed externally as arrays or tables does not mean that they are represented in an implementation as linear lists of objects. Any collection of same-typed objects is viewed as an array, even though it might be stored internally in some other format, for example, as a hash table.


   A HEMS query consists of a sequence of ASN.1 objects, interpreted by
   a simple stack-based interpreter.  [Although we define the query
   language in terms of the operations of a stack machine, the language
   does not require an implementation to use a stack machine.  This is a
   well-understood model, and is easy to implement.]  One ASN.1 tag is
   reserved for operation codes; all other tags indicate data that will
   eventually be used by an operation.  These objects are pushed onto
   the stack when received.  Opcodes are immediately executed and may
   remove or add items to the stack.  Because ASN.1 itself provides
   tags, very little needs to be done to the incoming ASN.1 objects to
   make them suitable for use by the query interpreter.

Each ASN.1 object in a query will fit into one of the following categories:

   Opcode    An opcode tells the query interpreter to perform an action.
             They are described in detail in section 8.  Opcodes are
             represented by an application-specific type whose value
             determines the operation.
   Template  These are objects that name one or more items in the data
             tree.  Named items may be either simple items (leaf nodes)
             or entire dictionaries, in which case the entire subtree
             "underneath" the dictionary is understood.  Templates are
             used to select specific data to be retrieved from the data
             tree.  A template may be either simple or structured,
             depending upon what it is naming.  A template only names
             the data -- there are no values contained in it.  Therefore
             the leaf objects in a template will all have a length of
             Examples of very simple templates are:
                 name()   or   System{}
             Each of these is just one ASN.1 data object, with zero
             length.  The first names a single data item in the "System"
             dictionary (and must appear in that context), and the
             second names the entire "System" dictionary.  A more
             complex template such as:
                 Interfaces{ InterfaceData{ address, netMask, ARP } }
             names two simple data items and a dictionary, iterated over
             all occurrences of "InterfaceData" within the Interfaces
   Path      A path is a special case of a template that names only a
             single node in the tree.  It specifies a path down into the
             dictionary tree and names exactly one node in the
             dictionary tree.
   Value     These are used to give data values when needed in a query,
             for example, when changing a value in the data tree.  A
             value can be thought of as either a filled-in template or
             as the ASN.1 representation some part of the data tree.
   Filter    A boolean expression that can be executed in the context of
             a particular dictionary that is used to select or not
             select items in the dictionary.  The expressions consist of
             the primitives "equal", "greater-or-equal",
             "less-or-equal", and "present" possibly joined by "and",
             "or", and "not".  (See section 8.6.)

Values, Paths, and Templates usually have names in the context- dependent class, except for a few special cases, which are in the application-specific class.


The data returned to the monitoring entity is a sequence of ASN.1 data items. Conceptually, the reply is a subset of the data tree, where the query selects which portions are to be included. This is exactly true for data retrieval requests, and essentially true for data modification requests -- the reply contains the data after it has been modified. The key point is that the data in a reply represents the state of the data tree immediately after the query was executed.

The sequence of the data is determined by the sequence of query language operations and the order of data items within Templates and Values given as input to these operations. If a query requests data from two of the top-level dictionaries in the data tree, by giving two templates such as:

          System{ name, interfaces }
                  InterfaceData { address, netMask, mtu }

then the response will consist of two ASN.1 data objects, as follows:

          System {
                  name("system name"),
          Interfaces {
                  InterfaceData { address(,
                  InterfaceData { address(,

With few exceptions, each of the data items in the hierarchy is named in the context-specific ASN.1 type space. Because of this, the returned objects must be fully qualified. For example, the name of the entity must always be returned encapsulated inside an ASN.1 object for "System". If it were not, there would be no way to tell if the object that was returned was "name" inside the "System" dictionary or "address" inside the "interfaces" dictionary (assuming in this case that "name" and "address" were assigned the same integer as their ASN.1 tags).

Having fully-qualified data simplifies decoding of the data at the receiving end and allows the tags to be locally chosen. Definitions for tags within routing tables won't conflict with definitions for tags within interfaces. Therefore, the people doing the name assignments are less constrained. In addition, most of the identifiers will be fairly small integers, which is an advantage because ASN.1 can fit tag numbers up to 30 in a one-octet tag field. Larger numbers require a second octet.

If data is requested that doesn't exist, either because the tag is not defined, or because an implementation doesn't provide that data (such as when the data is optional), the response will contain an ASN.1 object that is empty. The tag will be the same as in the query, and the object will have a length of zero.

The same response is given if the requested data does exist, but the invoker of the query does not have authorization to access it. See section 10 for more discussion of authorization mechanisms.

This allows completely generic queries to be composed without regard to whether the data is defined or implemented at all of the entities that will receive the query. All of the available data will be returned, without generating errors that might otherwise terminate the processing of the query.


The query language is designed to be expressive enough to write useful queries with, yet simple enough to be easy to implement. The query processor should be as simple and fast as possible, in order to avoid placing a burden on the monitored entity, which may be a critical node such as a gateway.

Although queries are formed in a flexible way using what we term a "language", this is not a programming language. There are operations that operate on data, but most other features of programming languages are not present. In particular:

- Programs are not stored in the query processor.

  • The only form of temporary storage is a stack, of limited depth.
  • There are no subroutines.
  • There are no explicit control structures defined in the language.

The central element of the language is the stack. It may contain templates, (and therefore paths), values, and filters taken from the query. In addition, it can contain dictionaries (and therefore arrays) from the data tree. At the beginning of a query, it contains one item, the root dictionary.

The overall operation consists of reading ASN.1 objects from the input stream. All objects that aren't opcodes are pushed onto the stack as soon as they are read. Each opcode is executed immediately and may remove items from the stack, may generate ASN.1 objects and send them to the output stream, and may leave items on the stack. Because each input object is dealt with immediately, portions of the response may be generated while the query is still being received.

In the descriptions below, operator names are in capital letters, preceded by the arguments used from the stack and followed by results left on the stack. For example:

   OP                             a b   OP   a t
             means that the OP operator takes <a> and <b> off of the
             stack and leaves <t> on the stack.  Most of the operators
             in the query language leave the first operand (<a> in this
             example) on the stack for future use.
   If both <a> and <b> were received as part of the query (as opposed to
   being calculated by previous operations), then this part of the query
   would have consisted of the sequence:
   So, like other stack-based languages, the arguments and operators
   must be presented in postfix order, with an operator following its

Here is a summary of all of the operators defined in the query language. Most of the operators can take several different sets of operands and behave differently based upon the operand types. Details and examples are given later.

   BEGIN                   dict1 path   BEGIN   dict1 dict
                    array path filter   BEGIN   array dict
             Move down in the data tree, establishing a context for
             future operations.
   END                           dict   END   --
             Undo the most recent BEGIN.
   GET                           dict   GET   dict
                        dict template   GET   dict
                array template filter   GET   array
             Retrieve data from the data tree.


                                 dict   GET-ATTRIBUTES   dict
                        dict template   GET-ATTRIBUTES   dict
                array template filter   GET-ATTRIBUTES   array
             Retrieve attribute information about data in the data tree.
   GET-RANGE   dict path start length   GET-RANGE   dict
             Retrieve a subrange of an OctetString.  Used for reading
   SET                     dict value   SET   dict
                   array value filter   SET   array
             Change values in the data tree, possibly performing control
   CREATE                 array value   CREATE   dict
             Create new table entries.
   DELETE                array filter   DELETE   array
             Delete table entries.

These operators are defined so that it is impossible to generate an invalid query response. Since a response is supposed to be a snapshot of a portion (or portions) of the data tree, it is important that only data that is actually in the tree be put in the response. Two features of the language help guarantee this:

  • Data is put in the response directly from the tree (by GET-*). Data does not go from the tree to the stack and then into the response.
  • Dictionaries on the stack are all derived from the initial, root dictionary. The operations that manipulate dictionaries (BEGIN and END) also update the response with the new location in the tree.

8.1 Moving Around in the Data Tree

The initial point of reference in the data tree is the root. That is, operators name data starting at the root of the tree. It is useful to be able to move to some other dictionary in the tree and then name data from that point. The BEGIN operator moves down in the tree and END undoes the last unmatched BEGIN.

BEGIN is used for two purposes:

  • By moving to a dictionary closer to the data of interest, the name of the data can be shorter than if the full name (from the root) were given.
  • It is used to establish a context for filtered operations to operate in. Filters are discussed in section 8.6.
   BEGIN                   dict1 path   BEGIN    dict1 dict
             Follow <path> down the dictionary starting from <dict1>.
             Push the final dictionary named by <path> onto the stack.
             <path> must name a dictionary (not a leaf node).  At the
             same time, produce the beginning octets of an ASN.1 object
             corresponding to the new dictionary.  It is up to the
             implementation to choose between using the "indefinite
             length" representation or the "definite length" form and
             going back and filling the length in later.
   END                           dict   END   --
             Pop <dict> off of the stack and terminate the open ASN.1
             object(s) started by the matching BEGIN.  Must be paired
             with a BEGIN.  If an END operation pops the root dictionary
             off of the stack, the query is terminated.

<path> must point to a regular dictionary. If any part of it refers to a non-existent node, if it points to a leaf node, or if it refers to a node inside an array-type dictionary, then it is in error, and the query is terminated immediately.

An additional form of BEGIN, which takes a filter argument, is described later.

8.2 Retrieving Data

The basic model that all of the data retrieval operations follow is that they take a template and fill in the leaf nodes of the template with the appropriate data values.

   GET                  dict template   GET   dict
             Emit an ASN.1 object with the same "shape" as the given
             template, except with values filled in for each node.  The
             first ASN.1 tag of <template> should refer to an object in
             <dict>.  If a dictionary tag is supplied anywhere in
             <template>, the entire dictionary contents are emitted to
             the response.  Any items in the template that are not in
             <dictionary> (or its components) are represented as objects
             with a length of zero.
                                 dict   GET   dict
             If there is no template, get all of the items in the
             dictionary.  This is equivalent to providing a template
             that lists all of the items in the dictionary.

An additional form of GET, which takes a filter argument, is described later.

Here is an example of using the BEGIN operator to move down the data tree to the TCP dictionary and then using the GET operator to retrieve 5 data values from the TCP Stats dictionary:

       IPTransport{ TCP } BEGIN
       Stats{ octetsIn, octetsOut, inputPkts, outputPkts, badtag } GET

This might return:

       IPTransport{ TCP
           Stats{ octetsIn(13255), octetsOut(82323),
                  inputPkts(9213), outputPkts(12425), badtag() }

"badtag" is a tag value that is undefined. No value is returned for it, indicating that there is no data value associated with it.

8.3 Data Attributes

Although ASN.1 "self-describes" the structure and syntax of the data, it gives no information about what the data means. For example, by looking at the raw data, it is possible to tell that an item is of type [context 5] and is 4 octets long. That does not tell how to interpret the data (is this an integer, an IP address, or a 4- character string?) or what the data means (IP address of what?). Even if the data were "tagged", in ASN.1 parlance, that would only give the base type (e.g., IP-address or counter) and not the meaning.

Most of the time, this information will come from RFC-1024, which defines the ASN.1 tags and their precise meaning. When extensions have been made, it may not be possible to get documentation on the extensions. (Extensions are discussed in section 9.)

The GET-ATTRIBUTES operator is similar to the GET operator, but returns a set of attributes describing the data rather than the data itself. This information is intended to be sufficient to let a human understand the meaning of the data and to let a sophisticated application treat the data appropriately. Such an application could use the attribute information to format the data on a display and decide whether it is appropriate to subtract one sample from another.

Some of the attributes are textual descriptions to help a human understand the nature of the data and provide meaningful labels for it. Extensive descriptions of standard data are optional, since they are defined in RFC-1024. Complete descriptions of extensions must be available, even if they are documented in a user's manual. Network firefighters may not have a current manual handy when the network is broken.

The format of the attributes is not as simple as the format of the data itself. It isn't possible to use the data's tag, since that would look exactly like the data itself. The format is:

       Attributes ::= [APPLICATION 3] IMPLICIT SEQUENCE {
               tagASN1         [0] IMPLICIT INTEGER,
               valueFormat     [1] IMPLICIT INTEGER,
               longDesc        [2] IMPLICIT IA5String OPTIONAL,
               shortDesc       [3] IMPLICIT IA5String OPTIONAL,
               unitsDesc       [4] IMPLICIT IA5String OPTIONAL,
               precision       [5] IMPLICIT INTEGER OPTIONAL,
               properties      [6] IMPLICIT BITSTRING OPTIONAL,
               valueSet        [7] IMPLICIT SET OF valueDesc OPTIONAL

The GET-ATTRIBUTES operator is similar to the GET operator. The major difference is that dictionaries named in the template do not elicit data for the entire subtree.


                        dict template   GET-ATTRIBUTES   dict
             Emit a single ASN.1 Attributes object for each of the
             objects named in <template>.  For each of these, the
             tagASN1 field will be set to the corresponding tag from the
             template.  The rest of the fields are set as appropriate
             for the data object.  Any items in the template that are
             not in <dictionary> (or its components) elicit an
             Attributes object with a valueFormat of NULL, and no other
             descriptive information.


                                 dict   GET-ATTRIBUTES   dict
             If there is no template, emit Attribute objects for all of
             the items in the dictionary.  This is equivalent to
             providing a template that lists all of the items in the
             dictionary.  This allows a complete list of a dictionary's
             contents to be obtained.

An additional form of GET-ATTRIBUTES, which takes a filter argument, is described later.

Here is an example of using the GET-ATTRIBUTES operator to request attributes for three objects in the System dictionary:

       System{ name, badtag, clock-msec } GET-ATTRIBUTES

"badtag" is some unknown tag. The result might be:



                       longDesc("The primary hostname."),
                       tagASN1(badtag), valueFormat(NULL)
                       longDesc("milliseconds since boot"),
                       shortDesc("uptime"), unitsDesc("ms")

Note that in this example "name" and "clock-msec" are integer values for the ASN.1 tags for the two data items. "badtag" is an integer value that has no corresponding name in this context.

There will always be exactly as many Attributes items in the result as there are objects in the template. Attributes objects for items which do not exist in the entity will have a valueFormat of NULL and none of the optional elements will appear.

       [ A much cleaner method would be to store the attributes as
       sub-components of the data item of interest.  For example,
           System{ clock-msec }  GET
       would normally just get the value of the data.  Asking for an
       additional layer down the tree would now get its attributes:
           System{ clock-msec{ shortDesc, unitsDesc }  GET
       would get the named attributes.  (The attributes would be
       named with application-specific tags.)  Unfortunately, ASN.1
       doesn't provide a notation to describe this type of
       organization.  So, we're stuck with the GET-ATTRIBUTES
       operator.  However, if a cleaner organization were possible,
       this decision would have been made differently. ]

8.4 Examining Memory

Even with the ability to symbolically access all of this information in an entity, there will still be times when it is necessary to get to very low levels and examine memory, as in remote debugging operations. The building blocks outlined so far can easily be extended to allow memory to be examined.

Memory is modeled as an ordinary object in the data tree, with an ASN.1 representation of OctetString. Because of the variety of addressing architectures in existence, the conversion from the internal memory model to OctetString is very machine-dependent. The only simple case is for byte-addressed machines with 8 bits per byte.

Each address space in an entity is represented by one "memory" data item. In a one-address-space situation, this dictionary will probably be in "System" dictionary. If each process has its own address space, then one "memory" item might exist for each process. Again, this is very machine-dependent.

Although the GET-RANGE operator is provided primarily for the purpose of retrieving the contents of memory, it can be used on any object whose base type is OctetString.

   GET-RANGE   dict path start length   GET-RANGE   dict
             Get <length> elements of the OctetString, starting at
             <start>.  <start> and <length> are both ASN.1 INTEGER type.
             <path>, starting from <dict>, must specify a node
             representing memory, or some other OctetString.

The returned data may not be <length> octets long, since it may take more than one octet to represent one memory location.

   Memory items in the data tree are special in that they will not
   automatically be returned when the entire contents of a dictionary
   are requested.  e.g., If memory is part of the "System" dictionary,
   then the query
       System GET
   will emit the entire contents of the System dictionary, but not the
   memory item.

8.5 Control Operations: Modifying the Data Tree

All of the operators defined so far only allow data in an entity to be retrieved. By replacing the template argument used in the GET operators with a value, data in the entity can be changed. Very few items in the data tree can be changed; those that can are noted in RFC-1024.

Values in the data tree can modified in order to change configuration parameters, patch routing tables, etc. Control functions, such as bringing an interface "down" or "up", do not usually map directly to changing a value. In such cases, an item in the tree can be defined to have arbitrary side-effects when a value is assigned to it. Control operations then consist of "setting" this item to an appropriate command code. Reading the value of such an item might return the current status. Again, details of such data tree items are given in RFC-1024.

This "virtual command-and-status register" model is very powerful, and can be extended by an implementation to provide whatever controls are needed. It has the advantage that the control function is associated with the controlled object in the data tree. In addition, no additional language features are required to support control functions, and the same operations used to locate data for retrieval are used to describe what is being controlled.

For all of the control and data modification operations, the fill- in-the-blank model used for data retrieval is extended: the response to an operation is the affected part of the data tree, after the operation has been executed. Therefore, for normal execution, SET and CREATE will return the object given as an argument, and DELETE will return nothing (because the affected portion was deleted).

   SET                     dict value   SET   dict
             Set the value(s) of data in the entity to the value(s)
             given in <value>.  The result will be the value of the data
             after the SET.  Attempting to set a non-settable item will
             not produce an error, but will yield a result in the reply
             different from what was sent.
   CREATE                 array value   CREATE   dict
             Insert a new entry into <array>.  Depending upon the
             context, there may be severe restrictions about what
             constitutes a valid <value>.  The result will be the actual
             item added to the <array>.  Note that only one item can be
             added per CREATE operation.
   DELETE                array filter   DELETE   array
             Delete the entry(s) in <array> that match <filter>.
             Filters are described later in this document.  Normally, no
             data items will be produced in the response, but if any of
             the items that matched the filter could not be deleted,
             they will be returned in the response.

An additional form of SET, which takes a filter argument, is described later.

   Here is an example of attempting to use SET to change the number of
   interfaces in an entity:
       System{ interfaces(5) } SET
   Since that is not a settable parameter, the result would be:
       System{ interfaces(2) }
   giving the old value.

Here is an example of how CREATE would be used to add a routing table entry for net for

       IPRouting BEGIN                   -- get dictionary
       Entries{ DestAddr(, ... }    -- entry to insert
       END                                 -- finished with dict

The result would be what was added:

       IPRouting{ Entries{ DestAddr(, ... } }
   The results in the response of these operators is consistent of the
   global model of the response:  it contains a subset of the data in
   the tree immediately after the query is executed.

Note that CREATE and DELETE only operate on arrays, and then only on arrays that are specifically intended for it. For example, it is quite reasonable to add and remove entries from routing tables or ARP tables, both of which are arrays. However, it doesn't make sense to add or remove entries in the "Interfaces" dictionary, since the contents of that array is dictated by the hardware. For each array in the data tree, RFC-1024 indicates whether CREATE and DELETE are valid.

CREATE and DELETE are always invalid in non-array contexts. If DELETE were allowed on monitored data, then the deleted data would become unmonitorable to the entire world. Conversely, if it were possible to CREATE arbitrary dictionary entries, there would be no way to give such entries any meaning. Even with the data in place, there is nothing that would couple the data to the operation of the monitored entity. [Creation and deletion would also add considerable complication to an implementation, because without them, all of the data structures that represent the data tree are essentially static, with the exception of dynamic tables such as the ones mentioned, which already have mechanisms in place for adding and removing entries.]

8.6 Associative Data Access: Filters

   One problem that has not been dealt with was alluded to earlier: When
   dealing with array data, how do you specify one or more entries based
   upon some value in the array entries?  Consider the situation where
   there are several interfaces.  The data might be organized as:
       Interfaces {                            -- one per interface
               InterfaceData { address, mtu, netMask, ARP{...}, ... }
               InterfaceData { address, mtu, netMask, ARP{...}, ... }

If you only want information about one interface (perhaps because

   there is an enormous amount of data about each), then you have to
   have some way to name it.  One possibility would be to just number
   the interfaces and refer to the desired interface as
   for the third one.

But this is not sufficient, because interface numbers may change over time, perhaps from one reboot to the next. It is even worse when dealing with arrays with many elements, such as routing tables, TCP connections, etc. Large, changing arrays are probably the more common case, in fact. Because of the lack of utility of indexing in this context, there is no general mechanism provided in the language for indexing.

A better scheme is to select objects based upon some value contained in them, such as the IP address. The query language uses filters to select specific table entries that an operator will operate on. The operators BEGIN, GET, GET-ATTRIBUTES, SET, and DELETE can take a filter argument that restricts their operation to entries that match the filter.

A filter is a boolean expression that is executed for each element in an array. If an array entry "matches" the filter (i.e., if the filter produces a "true" result), then it is used by the operation. A filter expression is very restricted: it can only compare data contained in the array element and the comparisons are only against constants. Comparisons may be connected by AND, OR and NOT operators.

The ASN.1 definition of a filter is:

       Filter          ::= [APPLICATION 2] CHOICE {
                               present         [0] DataPath,
                               equal           [1] DataValue,
                               greaterOrEqual  [2] DataValue,
                               lessOrEqual     [3] DataValue,
                               and             [4] SEQUENCE OF Filter,
                               or              [5] SEQUENCE OF Filter,
                               not             [6] Filter
       DataPath        ::= ANY                 -- Path with no value
       DataValue       ::= ANY                 -- Single data value

This definition is similar to the filters used in the ISO monitoring protocol (CMIP) and was derived from that specification.

"DataPath" is the name of a single data item; "DataValue" is the value of a single data item. The three comparisons are all of the form "data OP constant", where "data" is the value from the tree, "constant" is the value from the filter expression, and "OP" is one of equal, greater-than-or-equal, or less-than-or-equal. The last operation, "present", tests to see if the named item exists in the data tree. By its nature, it requires no value, so only a path needs to be given.

   Here is an example of a filter that matches an Interface whose IP
   address is
       Filter{ equal{ address( } }
   Note that the name of the data to be compared is relative to the
   "InterfaceData" dictionary.

Each operator, when given a filter argument, takes an array (dictionary containing only one type of item) as its first argument. In the current example, this would be "Interfaces". The items in it are all of type "InterfaceData". This tag is referred to as the "iteration tag".

Before a filtered operation is used, BEGIN must be used to put the array (dictionary) on top of the stack, to establish it as the context that the filter iterates over. The general operation of a filtered operation is then:

  1. Iterate over the items in the array.

2. For each element in the array, execute the filter.

  1. If the filter succeeds, do the requested operation (GET/SET/etc.) on the matched element, using the template/value/path as input to the operation. At this point, the execution of the operation is the same as in the non-filtered case.

This is a model of operation; actual implementations may take advantage of whatever lookup techniques are available for the particular table (array) involved.

Therefore, there are three inputs to a filtered operation:

  1. The "current dictionary" on the stack. This is the array-type dictionary to be searched, set by an earlier BEGIN.
  1. A filter, to test each item in the array. Each path or value mentioned in the filter must be named in the context of an item in the array, as if it was the current dictionary. For example, in the case where a filtered operation iterates over the set of "InterfaceData" items in the "Interfaces" array, each value or path in the filter should name an item in the "InterfaceData" dictionary, such as "address".
  1. A template, path, or value associated with the operation to be performed. The leading ASN.1 tag in this must match the iteration tag. In the current example where the filter is searching the "Interfaces" dictionary, the first tag in the template/tag/value must be "InterfaceData".

The operators which take filters as arguments are:

   BEGIN            array path filter   BEGIN   array dict
             Find a dictionary in <array> that matches <filter>.  Use
             that as the starting point for <path> and push the
             dictionary corresponding to <path> onto the stack.  If more
             than one dictionary matches <filter>, then any of the
             matches may be used.  This specification does not state how
             the choice is made.  At least one dictionary must match; it
             is an error if there are no matches.  (Perhaps it should be
             an error for there to be multiple matches; actual
             experience is needed to decide.)
   GET          array template filter   GET   array
             For each item in <array> that matches <filter>, fill in the
             template with values from the data tree and emit the
             result.  The first tag of <template> must be equal to the
             iteration tag.  Selected parts of matched items are emitted
             based upon <template>, just as in a non-filtered GET


                array template filter   GET-ATTRIBUTES   array
             Same as GET, except emit attributes rather than data
   SET             array value filter   SET   array
             Same as GET, except set the values in <value> rather than
             retrieving values.  Several values in the data tree will be
             changed if the filter matches more than one item in the
   DELETE                array filter   DELETE   array
             Delete the entry(s) in <array> that match <filter>.

Notes about filter execution:

  • Expressions are executed by inorder tree traversal.
  • Since the filter operations are all GETs and comparisons, there are no side-effects to filter execution, so an implementation is free to execute only as much of the filter as required to produce a result (e.g., don't execute the rest of an AND if the first comparison turns out to be false).
      - It is not an error for a filter to test a data item that
        isn't in the data tree.  In this situation, the comparison
        just fails (is false).  This means that filters don't need
        to test for the existence of optional data before
        attempting to compare it.

Here is an example of how filtering would be used to obtain the input and output packet counts for the interface with IP address

       Interfaces BEGIN                      -- dictionary
       InterfaceData{ pktsIn, pktsOut }      -- template
       Filter{ equal{ address( } }
       END                                   -- finished with dict

The returned value would be something like:

       Interfaces{                             -- BEGIN
         InterfaceData{ pktsIn(1345134), pktsOut(1023729) }
                                               -- GET
       }                                       -- END

The annotations indicate which part of the response is generated by the different operators in the query.

Here is an example of accessing a table contained within some other table. Suppose we want to get at the ARP table for the interface with IP address and retrieve the entire ARP entry for the host with IP address In order to retrieve a single entry in the ARP table (using a filtered GET), a BEGIN must be used to get down to the ARP table. Since the ARP table is contained within the Interfaces dictionary (an array), a filtered BEGIN must be used.

       Interfaces BEGIN                        -- dictionary
       InterfaceData{ ARP }                    -- path
       Filter{ equal{ address( } }    -- filter
       BEGIN                                   -- filtered BEGIN
       -- Now in ARP table for; get entry for
       addrMap                                 -- whole entry
       Filter{ equal{ ipAddr( } }    -- filter
       GET                                     -- filtered GET

The result would be:

       Interfaces{                             -- first BEGIN
         InterfaceData{ ARP{                   -- second BEGIN
           addrMap{ ipAddr(, physAddr(..) } -- from GET
         } }                                   -- first END
       }                                       -- second END

Note which parts of the output are generated by different parts of the query.

Here is an example of how the SET operator would be used to shut down the interface with ip-address by changing its status to "down".

       Interfaces BEGIN                    -- get dictionary
       Interface{ Status(down) }           -- value to set
       Filter{ equal{ IP-addr( } }

If the SET is successful, the result would be:

       Interfaces{                             -- BEGIN
           Interface{ Status(down) }           -- from SET
       }                                       -- END

8.7 Terminating a Query

A query is implicitly terminated when there are no more ASN.1 objects to be processed by the interpreter. For a perfectly-formed query, the interpreter would be back in the state it was when it started: the stack would have only the root dictionary on it, and all of the ASN.1 objects in the result would be terminated.

If there are still "open" ASN.1 objects in the result (caused by leaving ENDs off of the query), then these are closed, as if a sufficient number of ENDs were provided. This condition would be indicated by the existence of dictionaries other than the root dictionary on the stack.

If an extra END is received that would pop the root dictionary off of the stack, the query is terminated immediately. No error is generated.


There are two ways to extend the set of values understood by the query language. The first is to register the data and its meaning and get an ASN.1 tag assigned for it. This is the preferred method because it makes that data specification available for everyone to use.

The second method is to use the VendorSpecific application type to "wrap" the vendor-specific data. Wherever an implementation defines data that is not in RFC-1024, the "VendorSpecific" tag should be used to label a dictionary containing the vendor-specific data. For example, if a vendor had some data associated with interfaces that was too strange to get standard numbers assigned for, they could, instead represent the data like this:

interfaces {

                  interface {
                          in-pkts, out-pkts, ...
                          VendorSpecific { ephemeris, declination }

In this case, ephemeris and declination correspond to two context- dependent tags assigned by the vendor for their non-standard data.

If the vendor-specific method is chosen, the private data MUST have descriptions available through the GET-ATTRIBUTES operator. Even with this descriptive ability, the preferred method is to get standard numbers assigned if possible.


This specification does not state what type of authorization system is used, if any. Different systems may have needs for different mechanisms (authorization levels, capability sets, etc.), and some systems may not care about authorization at all. The only effect that an authorization system has on a query is to restrict what data items in the tree may be retrieved or modified.

Therefore, there are no explicit query language features that deal with protection. Instead, protection mechanisms are implicit and may make some of the data invisible (for GET) or non-writable (for SET):

  • Each query runs with some level of authorization or set of capabilities, determined by its environment (HEMS and the HEMP header).
  • Associated with each data item in the data tree is some sort of test to determine if a query's authorization should grant it access to the item.

Authorization tests are only applied to query language operations that retrieve information (GET, GET-ATTRIBUTES, and GET-RANGE) or modify it (SET, CREATE, DELETE). An authorization system must not affect the operation of BEGIN and END. In particular, the authorization must not hide entire dictionaries, because that would make a BEGIN on such a dictionary fail, terminating the entire query.


If some particular information is requested but is not available, it will be returned as "no-value" by giving the ASN.1 length as 0.

When there is any other kind of error, such as having improper arguments on the top of the stack or trying to execute BEGIN when the path doesn't refer to a dictionary, an ERROR object is emitted.

The contents of this object identify the exact nature of the error:

                  errorCode       INTEGER,
                  errorInstance   INTEGER,
                  errorOffset     INTEGER
                  errorDescription IA5String,
                  errorOp         INTEGER,

errorCode identifies what the error was, and errorInstance is an implementation-dependent code that gives a more precise indication of where the error occured. errorOffset is the location within the query where the error occurred. If an operation was being executed, errorOp contains its operation code, otherwise zero. errorDescription is a text string that can be printed that gives some description of the error. It will at least describe the errorCode, but may also give details implied by errorInstance. Detailed definitions of all of the fields are given in appendix I.2.

Since there may be several unterminated ASN.1 objects in progress at the time the error occurs, each one must be terminated. Each unterminated object will be closed with a copy of the ERROR object. Depending upon the type of length encoding used for this object, this will involve filling the value for the length (definite length form) or emitting two zero octets (indefinite length form). After all objects are terminated, a final copy of the ERROR object will be emitted. This structure guarantees that the error will be noticed at every level of interpretation on the receiving end.

In summary, if there was an error before any ASN.1 objects were generated, then the result would simply be:


If a couple of ASN.1 objects were unterminated when the error occurred, the result might look like:

            interface { name(...) type(...) error{...} }

It would be possible to define a "WARNING" object that has a similar (or same) format as ERROR, but that would be used to annotate responses when a non-fatal "error" occurs, such as attempting to SET/CREATE/DELETE and the operation is denied. This would be an additional complication, and we left it out in the interests of simplicity.


A query consists of a sequence of ASN.1 objects, as follows:

Query := IMPLICIT SEQUENCE of QueryElement;

QueryElement ::= CHOICE {


Operation and Filter are defined below. The others are:

       Template        ::= any
       Path            ::= any
       InputValue      ::= any

These three are all similar, but have different restrictions on their structure:

   Template        Specifies a portion of the tree, naming one or more
                   values, but not containing any values.
   Path            Specifies a single path from one point in the tree to
                   another, naming exactly one value, but not containing
                   a value.
   InputValue      Gives a value to be used by a query language

A query response consists of a sequence of ASN.1 objects, as follows:

Response := IMPLICIT SEQUENCE of ResponseElement;

ResponseElement ::= CHOICE {


Error is defined below. The others are:

       ResultValue     ::= any

ResultValue is similar to Template, above:

   ResultValue     Specifies a portion of the tree, naming and
                   containing one or more values.

The distinctions between these are elaborated in section 6.

I.1 Operation Codes

Operation codes are all encoded in a single application-specific type, whose value determines the operation to be performed. The definition is:

       Operation ::= [APPLICATION 1] IMPLICIT INTEGER {

I.2 Error Returns

An Error object is returned within a reply when an error is encountered during the processing of a query. Note that the definition this object is similar to that of the HEMP protocol error structure. The error codes have been selected to keep the code spaces distinct between the two. This is intended to ease the processing of error messages. See section 11 for more information.

               errorCode       INTEGER,
               errorInstance   INTEGER,
               errorOffset     INTEGER
               errorDescription IA5String,
               errorOp         INTEGER,

The fields are defined as follows:

   errorCode       Identifies the general cause of the error.
   errorInstance   An implementation-dependent code that gives a more
                   precise indication of where the error occured in the
                   query processor.  This is most useful when internal
                   errors are reported.
   errorOffset     The location within the query where the error was
                   detected.  The first octet of the query is numbered
   errorOp         If an operation was being executed, this contains its
                   operation code, otherwise zero.


A text string that can be printed that gives some description of the error. It will at least describe the errorCode, but may also give details implied by errorInstance.

Some errors are associated with the execution of specific operations, and others with the overall operation of the query interpreter. The errorCodes are split into two groups.

The first group deals with overall interpreter operation. Except for

"unknown operation", these do not set errorOp.

   100             Other error.
                   Any error not listed below.
   101             Format error.
                   An error has been detected in the format of the input
                   stream, preventing further interpretation of the
   102             System error.
                   The query processor has failed in some way due to an
                   internal error.
   103             Stack overflow.
                   Too many items were pushed on the stack.
   104             Unknown operation.
                   The operation code is invalid.  errorOp is set.

The second group is errors that are associated with the execution of particular operations. errorOp will always be set for these.

   200             Other operation error.
                   Any error, associated with an operation, not listed
   201             Stack underflow.
                   An operation expected to see some number of operands
                   on the stack, and there were fewer items on the
   202             Operand error.
                   An operation expected to see certain operand types on
                   the stack, and something else was there.
   203             Invalid path for BEGIN.
                   A path given for BEGIN was invalid, because some
                   element in the path didn't exist.
   204             Non-dictionary for BEGIN.
                   A path given for BEGIN was invalid, because the given
                   node was a leaf node, not a dictionary.
   205             BEGIN on array element.
                   The path specified an array element.  The path must
                   point at a single, unique, node.  A filtered BEGIN
                   should have been used.
   206             Empty filter for BEGIN.
                   The filter for a BEGIN didn't match any array
   207             Filtered operation on non-array.
                   A filtered operation was attempted on a regular
                   dictionary.  Filters can only be used on arrays.
   208             Index out of bounds.
                   The starting address or length for a GET-RANGE
                   operation went outside the bounds for the given
   209             Bad object for GET-RANGE.
                   GET-RANGE can only be applied to objects whose base
                   type is OctetString.

This list is probably not quite complete, and would need to be extended, based upon implementation experience.

I.3 Filters

Many of the operations can take a filter argument to select among elements in an array. They are discussed in section 8.6.

        Filter          ::= [APPLICATION 2] CHOICE {
                               present         [0] DataPath,
                               equal           [1] DataValue,
                               greaterOrEqual  [2] DataValue,
                               lessOrEqual     [3] DataValue,
                               and             [4] SEQUENCE OF Filter,
                               or              [5] SEQUENCE OF Filter,
                               not             [6] Filter
       DataPath        ::= ANY                 -- Path with no value
       DataValue       ::= ANY                 -- Single data value

A filter is executed by inorder traversal of its ASN.1 structure.

The basic filter operations are:

   present         tests for the existence of a particular data item in
                   the data tree
   equal           tests to see if the named data item is equal to the
                   given value.
   greaterOrEqual  tests to see if the named data item is greater than
                   or equal to the given value.
   lessOrEqual     tests to see if the named data item is less than or
                   equal to the given value.

These may be combined with "and", "or", and "not" operators to form arbitrary boolean expressions. The "and" and "or" operators will take any number of terms. Terms are only evaluated up to the point where the outcome of the expression is determined (i.e., an "and" term's value is false or an "or" term's value is true).

I.4 Attributes

One or more Attributes structure is returned by the GET-ATTRIBUTES operator. This structure provides descriptive information about items in the data tree. See the discussion in section 8.3.

       Attributes ::= [APPLICATION 3] IMPLICIT SEQUENCE {
               tagASN1         [0] IMPLICIT INTEGER,
               valueFormat     [1] IMPLICIT INTEGER,
               longDesc        [2] IMPLICIT IA5String OPTIONAL,
               shortDesc       [3] IMPLICIT IA5String OPTIONAL,
               unitsDesc       [4] IMPLICIT IA5String OPTIONAL,
               precision       [5] IMPLICIT INTEGER OPTIONAL,
               properties      [6] IMPLICIT BITSTRING OPTIONAL,
               valueSet        [7] IMPLICIT SET OF valueDesc OPTIONAL
       valueDesc ::= IMPLICIT SEQUENCE {
               value           [0] ANY,        -- Single data value
               desc            [1] IA5String

The meanings of the various attributes are given below.

   tagASN1         The ASN.1 tag for this object.  This attribute is
   valueFormat     The underlying ASN.1 type of the object (e.g.,
                   SEQUENCE or OCTETSTRING or Counter).  This is not
                   just the tag number, but the entire tag, as it would
                   appear in an ASN.1 object.  As such, it includes the
                   class, which should be either UNIVERSAL or
                   APPLICATION.  Applications receiving this should

ignore the constructor bit. This attribute is required.

   longDesc        A potentially lengthy text description which fully
                   defines the object.  This attribute is optional for
                   objects defined in this memo and required for
                   entity-specific objects.
   shortDesc       A short mnemonic string of less than 15 characters,
                   suitable for labeling the value on a display.  This
                   attribute is optional.
   unitsDesc       A short string used for integer values to indicate
                   the units in which the value is measured (e.g., "ms",
                   "sec", "pkts", etc.).  This attribute is optional.
   precision       For Counter objects, the value at which the Counter
                   will roll-over.  Required for all Counter objects.
   properties      A bitstring of boolean properties of the object.  If
                   the bit is on, it has the given property.  This
                   attribute is optional.  The bits currently defined
                   0   If true, the difference between two values of
                       this object is significant.  For example, the
                       changes of a packet count is always significant,
                       it always conveys information.  In this case, the
                       0 bit would be set.  On the other hand, the
                       difference between two readings of a queue length
                       may be meaningless.
                   1   If true, the value may be modified with SET,
                       CREATE, and DELETE.  Applicability of CREATE and
                       DELETE depends upon whether the object is in an
                   2   If true, the object is a dictionary, and a BEGIN
                       may be used on it.  If false, the object is leaf
                       node in the data tree.
                   3   If true, the object is an array-type dictionary,
                       and filters may be used to traverse it.  (Bit 2
                       will be true also.)
   valueSet        For data that is defined as an ASN.1 CHOICE type (an
                   enumerated type), this gives descriptions for each of
                   the possible values that the data object may assume.

Each valueDesc is a <value,description> pair. This information is especially important for control items, which are very likely to appear in VendorSpecific dictionaries, exactly the situation where descriptive information is needed.

I.5 VendorSpecific

See the discussion in section 9.

       VendorSpecific          ::= [APPLICATION 4] IMPLICIT SET
                                       of ANY


Although it is not normally in the spirit of RFCs to define an implementation, the authors feel that some suggestions will be useful to implementors of the query language. This list is not meant to be complete, but merely to give some hints about how the authors imagine that the query processor might be implemented efficiently.

  • It should be understood that the stack is of very limited depth. Because of the nature of the query language, it can get only about 4 entries (for arguments) plus the depth of the tree (up to one BEGIN per level in the tree). This comes out to about a dozen entries in the stack, a modest requirement.
  • The stack is an abstraction -- it should be implemented with pointers, not by copying dictionaries, etc.
  • An object-oriented approach should make implementation fairly easy. Changes to the "shape" if the data items (which will certainly occur, early on) will also be easier to make.
  • Only a few "messages" need to be understood by objects. By having pointers to action routines for each basic operation (GET,SET,...) associated with each node in the tree, common routines (e.g., emit a long integer located at address X) can be shared, and special routines (e.g., set the interface state for interface X) can be implemented in a common framework. Higher levels need know nothing about what data is being dealt with.
  • Most interesting objects are dictionaries, each of which can be implemented using pointers to the data and procedure "hooks" to perform specific operations such as GET, SET, filtering, etc.
  • The hardest part is actually extracting the data from existing TCP/IP implementations that weren't designed with detailed monitoring in mind. Query processors interfacing to a UNIX kernel will have to make many system calls in order to extract some of the more intricate structures, such as routing tables. This should be less of a problem if a system is designed with easy monitoring as a goal.

A Skeletal Implementation

This section gives a rather detailed example of the core of a query processor. This code has not been tested, and is intended only to give implementors ideas about how to tackle some aspects of query processor implementation with finesse, rather than brute force.

The suggested architecture is for each dictionary to have a "traverse" routine associated with it, which is called when any sort of operation has to be done on that dictionary. Most nodes will share the same traversal routine, but array dictionaries will usually have routines that know about whatever special lookup mechanisms are required.

Non-dictionary nodes would have two routines, "action", and "compare", which implement query language operations and filter comparisons, respectively. Most nodes would share these routines.

For example, there should be one "action" routine that does query language operations on 32-bit integers, and another that works on 16-bit integers, etc.

Any traversal procedure would take arguments like:

       traverse(node, mask, op, filter)
               Treenode        node;   /* generic node-in-tree */
               ASN             mask;   /* internal ASN.1 form */
               enum opset      op;     /* what to do */
               Filter          filter; /* zero if no filter */
       enum opset { begin, get, set, create, delete, geta,
                       c_le, c_ge, c_eq, c_exist };

The traversal procedure is called whenever anything must be done within a dictionary. The arguments are:

   node            the current dictionary.
   mask            is either the template, path, or value, depending
                   upon the operation being performed.  The top-level
                   identifier of this object will be looked up in the
                   context of <node>.
   op              is the operation to be performed, either one of the
                   basic operations, or a filter operation.
   filter          is the filter to be applied, or zero if none.  There
                   will be no filter when <op> is a filter-comparison

The general idea is that the traversal proc associated with a node has all of the knowledge about how to get around in this subtree encoded within it. Hopefully, this will be the only place this knowledge is coded. Here is a skeleton of the "standard" traversal proc, written mostly in C.

When the query processor needs to execute a "GET" operation, it would just call:

       traverse(current, template, GET, 0)

Notes about this example:

  • This traversal routine handles either query language operations (GET, SET, etc.) or low-level filter operations. Separate routines could be defined for the two classes of operations, but they do much of the same work.
  • Dictionary nodes have a <traversal> proc defined.
  • Leaf nodes have an <action> proc, which implement GET, SET, GET-ATTRIBUTES, CREATE, and DELETE, and a <compare> proc, which performs low-level filter comparisons.
  • In the generic routine, the filter argument is unused, because the generic routine isn't used for array dictionaries, and only array dictionaries use filters.
  • An ASN type contains the top level tag and a list of sub-components.
  • size(mask) takes an ASN.1 object and tells how many sub-items are in it. Zero means that this is a simple object.
  • lookup(node, tag) looks up a tag in the given (tree)node, returning a pointer to the node. If the tag doesn't exist

in that node, a pointer to a special node "NullItem" is

returned. NullItem looks like a leaf node and has procs

that perform the correct action for non-existent data.

  • This example does not do proper error handling, or ASN.1 generation, both of which would require additional code in this routine.
        *  For op = GET/SET/etc, return:
        *              true on error, otherwise false.
        *  When op is a filter operation, return:
        *              the result of the comparison.
       int std_traverse(node, mask, op, filter)
           Treenode    node;   /* current node */
           ASN         mask;   /* internal ASN.1 form */
           enum opset  op;     /* what to do */
           Filter      filter; /* unused in this routine */
           ASN         item;
           Treenode    target;
           boolean     rv = false;
           extern Treenode NullItem;
           if (filter != null) {
               return true;
           target = lookup(node, mask.tag);
           /*  We are at the leaf of the template/path/value.  */
           if (size(mask) == 0)
               switch (op)
               case BEGIN:
                   /*  non-existent node, or leaf node  */
                   if (target == NullItem || target.traverse == 0) {
                       return true;
                   else {
                       begin(node, mask.tag);
                       return false;

case GET: case SET: case GETA:

               case GETR:      case CREATE:    case DELETE:
                   /*  A leaf in the mask specifies entire directory.
                       For GET, traverse the entire subtree.  */
                   if (target.traverse)
                       if (op == GET) {
                           foreach subnode in target
                               /*  Need to test to not GET memory.  */
                               rv |= (*target.traverse)
                                       (target, subnode.tag, op, 0);
                           return rv;
                       else if (op == SET)     /*  no-op  */
                           return false;
                       else if (op != GETA) {
                           return true;
                   /*  We're at a leaf in both the mask and the tree.
                       Just execute the operation.
                   else {
                       if (op == BEGIN) {  /*  Can't begin on leaf  */
                           return true;
                           return (*target.action)(target, mask, op);
                   }  /* else */
               default:        /*  Comparison ops.  */
                   return (*, mask, op);
               }  /* switch */

/* We only get here if mask has structure. */

           /*  can't have multiple targets for BEGIN  */
           if (op == BEGIN && size(mask) != 1) {
               return true;
           /*  or for a single filter operation.  */
           if (op is comparison && size(mask) != 1) {
               return false;
           /*  Iterate over the components in mask  */
           foreach item in mask
               if (target.traverse)    /*  traverse subtree.  */
                   rv |= (*component.traverse)(component, item, op, 0);
               else                    /*  leaf node, at last.  */
                   if (op is comparison)
                       return (*, mask, op);
                       return (*target.action)(target, mask, op);
           } /* foreach */
           return rv;
       }  /* std_traverse */

Here is a bare skeleton of an array-type dictionary's traversal proc.

       int array_traverse(node, mask, op, filter)
           Treenode    node;   /* current node */
           ASN         mask;   /* internal ASN.1 form */
           enum opset  op;     /* what to do */
           Filter      filter; /* unused in this routine */
           Treenode    target;
           boolean     rv = false;
           extern Treenode NullItem;
           /*  Didn't find that key.  */
           if (mask.tag != this array's iteration tag)
               return false;
           if (op == BEGIN && filter == null) {
               return 1;
           /*  The implementation of this loop is the major trick!  */
           /*  Needs to stop after first filter success on BEGIN.  */
           foreach target in node {
               if (filter == null ||           /*  if no filter, or */
                   ExecFilter(target, filter)) /* if it succeeds  */
                   rv |= (target.traverse*)(target, mask, op, 0);
       }  /* array_traverse */

Object-oriented programming languages, such as C++, Modula, and Ada, are well suited to this style of implementation. There should be no particular difficulty with using a conventional language such as C or Pascal, however.


Copies of ISO Standard ASN.1 (Abstract Syntax Notation 1) are available from the following source. It comes in two parts; both are needed:

       IS 8824 -- Specification (meaning, notation)
       IS 8825 -- Encoding Rules (representation)

They are available from:

       Omnicom Inc.
       115 Park St, S.E.          (new address as of March, 1987)
       Vienna, VA  22180
       (703) 281-1135