Copyright © 1998 W3C (MIT, INRIA, Keio ), All Rights Reserved. W3C liability, trademark, document use and software licensing rules apply.
This is a W3C Working Draft for review by W3C members and other interested parties. It is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to use W3C Working Drafts as reference material or to cite them as other than "work in progress". A list of current W3C technical reports can be found at http://www.w3.org/TR.
This document has been produced as part of the W3C HTTP-ng Activity. This is work in progress and does not imply endorsement by, or the consensus of, either W3C or members of the HTTP-ng Protocol Design Working Group. We expect the document to evolve as we get more data from the Web Characterization Group describing the current state of the Web.
This document describes a binary `on-the-wire' protocol to be used when sending HTTP-ng operation invocations or terminations across a network connection. It is part of a suite of documents describing the HTTP-NG design and prototype implementation:
Please send comments on this specification to <www-http-ng-comments@w3.org>.
Two data description languages are used in this document. The first, called ISL, is an abstract language for defining data types and interfaces. It is described in the ILU manual. The second is a pseudo-C syntax. It should be interpreted as C data structure layouts without any automatic padding to size boundaries, and allowing arbitrary bit-size limits on structs and unions as well as on ints and enums. Each use of ISL and pseudo-C is marked as to which language is being used.
This protocol assumes a particular model of operation based on conventional RPC technology, with certain variations. The basic idea is that clients make use of services exported from a server by invoking operations on objects resident in that server. The client is connected to the server by a connection, which carries operation invocation requests from the client (known as the caller) to the server (known as the callee), and operation results from the callee back to the caller. Multiple connections can exist simultaneously between the same client and server. The connection has state associated with it, which allows the caller and callee to use shorthand notations for some of the data passed to the other party.
Two RPC messages are defined by this protocol: the Request, which is used by the caller to invoke an operation on the callee, and the Reply, which is used to transfer operation results from the callee to the caller. Every Reply message is associated with a particular Request message, but not every Request message has a Reply message associated with it. Connections are directional; operation invocation Requests always flow from the caller to the callee; Replies always flow from the callee to the caller. In addition to the RPC messages, several control messages are defined for this protocol. These control messages are used to improve the efficiency and robustness of the connection. They are intended to be generated and consumed by the implementation of the wire protocol, and should have no direct effect on the applications using the protocol.
A Request message indicates two important elements, the operation and the discriminant object, or discriminant; it also contains data values which are the input parameters to the operation. The model used here assumes that operations are grouped into sets, the elements of which have a well-defined ordering; each operation set is called an interface. It further assumes that an interface can be identified by a URN which also a UUID; and that each operation in an interface can be identified with the ordinal number of the operation within the ordering of the elements of the interface. It assumes that every discriminant object can be identified with an object ID, also a URN and UUID. It provides for the fact that, with most distributed object systems, all of the discriminants available at a particular server share a common prefix to their object ID; this is called the server ID. Note that this characteristic is not required, but the protocol provides an efficiency optimization for the case where it is true. In such a case, we call the portion of the object ID not contained in the server ID the instance handle. Each request has an implicit connection-specific serial number associated with it; serial numbers begin with the value one (1), and have a maximum value of 16777215. When the maximum serial number of a connection has been reached, the connection must be terminated, and further operations must be invoked over a new connection.
A Reply message indicates the termination status of the operation, provides information about synchronization, and may contain data values which are output parameters or `return values' from the operation. It contains an explicit serial number to indicate which Request it is a reply to. Replies may either indicate successful completion of the operation, or several different kinds of exceptional termination; if an exception is signalled, additional information is passed to indicate which of the possible exceptions for the operation was raised.
The model assumes that the messages are carried back and forth between the two parties by a transport subsystem. It requires that the transport subsystem be reliable, sequenced, and message-oriented. By reliable, we mean that after a message is handed to the transport, the transport will either deliver it to the other party, or will signal an error if its reliable delivery cannot be ascertained. By sequenced, we mean that the transport will deliver messages to the other party in the same order in which the sender handed them to the transport. By message-oriented, we mean that the transport will provide indication of the beginning and ending of the messages, without reference to any data encoded inside the message. An example of this type of transport would be the record marking defined in Internet RFC 1831 used with TCP/IP.
All values use `network standard' byte order, i.e. big-endian, because all Internet protocols use it. If in the future this becomes a problem for the Internet, this protocol will be affected by whatever solution is used to solve the problem in the wider Internet context. Note that the data marshalling format defined in Internet RFC 1832, which this protocol incorporates by reference, is also defined to be a big-endian protocol.
The marshalled form of each value begins on a 32-bit boundary. The marshalled form of each value is padded-after, if necessary, to the next 32-bit boundary. The padding bits may be either 0 or 1 in any combination.
Marshalling is via the XDR format specified in Internet RFC 1832. It could be argued that this format is inexcusably wasteful with certain value types, such as boolean (32 bits) or byte (32 bits), and that a 16-bit or 8-bit oriented format should be designed and used in its place. However, the argument of using an existing Internet standard for this purpose, rather than inventing a new one, is a strong one; a new format should only be defined if measurement of the overhead shows gross waste.
This protocol assumes that security provisions are made either at some level above it, typically in the application interfaces, or at some level below it, typically by use of a secure transport mechanism. It contains no protocol-level mechanisms for providing or assuring any of the concerns normally related to security.
Unlike some previous protocols, this protocol is session-oriented. That means that individual messages are sent in the context of a session, and are context-sensitive. This context-sensitivity allows session-wide compression. However, to support various kinds of marshalling architectures in implementations of this system, all marshalling can be done in a context-insensitive fashion, at the expense of sending additional bytes across the wire. However, unmarshalling implementations must always be capable of tracking and using context-sensitive information.
The following data structures are defined in pseudo-C:
typedef enum {
False = 0,
True = 1
} Boolean;
typedef enum {
InitializeConnection = 0,
TerminateConnection = 1,
DefaultCharset = 2
} ControlMsgType;
typedef enum {
Success = 0,
UserException = 1, /* occurred during operation */
SystemExceptionBefore = 2, /* occurred before beginning operation */
SystemExceptionAfter = 3 /* occurred after beginning operation */
} ReplyStatus;
typedef struct {
Boolean cached_disc : 1; /* True if cached object key */
union {
struct {
Boolean cache_key : 1; /* True if both sides cache it */
unsigned key_len : 13; /* length of key bytes */
} uncached_key;
unsigned cache_index : 14; /* cache index if cached */
} v;
} DiscriminantID;
typedef struct {
Boolean cached_op : 1; /* True if cached id */
union {
struct {
Boolean cache_operation : 1; /* True if should be cached */
unsigned method_id : 13; /* method index */
} uncached_op_info;
unsigned cache_index : 14; /* cache index if "cached_op" set */
} v;
} OperationID;
typedef enum {
MangledMessage = 0, /* bad protocol synchronization */
ProcessFinished = 1, /* sending party has `exitted' */
ResourceManagement = 2, /* transient close */
WrongCallee = 3, /* bad server ID received */
MaxSerialNumber = 4 /* the maximum serial number was used */
} TerminationCause;
typedef struct {
unsigned major : 4;
unsigned minor : 4;
} ProtocolVersion;
typedef unsigned Unused;
Only a few messages are defined. The InitializeConnection message
is used by the caller to verify that it has connected to the right server,
and that it is using the correct version of the wire protocol. The
DefaultCharset message allows both sides to independently define
a default value for string charsets. The Request message causes
an operation to be started on the remote server. The Reply message
is sent from the server to the client to inform it of the completion status
of the operation, and to convey any result values. The
TerminateConnection message allows either side to indicate graceful
shutdown of a connection.
This protocol uses a feature called an extension header to provide
for extensibility and tailorability. Features such as serialization contexts
or global thread identifiers may be implemented via this feature. An extension
header is an encapsulated value of the ISL type ExtensionHeader.
Each request message and reply message may contain a value of type
ExtensionHeaderList, which contains a number of extension headers.
The following ISL fragment decribes the types
ExtensionHeaderList and ExtensionHeader:
INTERFACE HTTP-ng-w3ng IMPORTS HTTP-ng END BRAND "http-ng.w3.org";
...
TYPE SimpleString = STRING LANGUAGE "i-default" LIMIT 0xFFFF;
TYPE CinfoString = STRING LANGUAGE "i-httpngcinfo" LIMIT 0xFFFF;
TYPE ExtensionHeader = RECORD
name : HTTP-ng.UUIDString,
value : PICKLE
END;
TYPE ExtensionHeaderList = SEQUENCE OF ExtensionHeader;
...
Request Message
Request header (pseudo-C):
typedef struct {
Boolean control_msg : 1; /* == FALSE */
Boolean ext_hdr_present : 1; /* True if ext hdr list present */
OperationID operation_id : 15; /* identifies operation */
DiscriminantID object_key : 15; /* identifies discriminant */
} RequestMsgHeader /* 4 bytes total */
The actual message consists of the following sections:
[ RequestMsgHeader ]
[ extension header list, if any ]
[ XDR string containing object type ID of object type defining
operation, if not cached ]
[ bytes of object_key, if not cached, padded to 4 byte boundary
]
[ explicit input parameter values, if any, padded to a 4 byte boundary ]
The operation_id contains either a connection-specific 14-bit
cache index, or a 13-bit method id (the zero-based ordinal position of the
method in the ISL declaration of the object type in which the operation is
defined) of the operation. If the method id is given, an additional value,
an XDR string value containing the object type ID of the object
type in which the operation is defined, is also passed. This means that this
protocol will not support interfaces in which object types have more than
8192 methods directly defined.
The object_key is either a 14-bit connection-specific cache index,
or the length of a variable length octet sequence of 8192 or fewer bytes
containing the service-point-relative name for the object (the
instance-handle of the URL). The object key value of { False,
False, 0 }, normally a zero byte variable length object key, is reserved
for use by the protocol. The object_key is marshalled onto the
transport as an XDR value of type fixed-length opaque data,
where the length is that specified in the v.key_len field of
the object_key.
Callers may reduce the size of messages by memoizing operation IDs and object
IDs that are passed in the connection. This is done by the caller setting
the cache_key (for object IDs) or cache_operation
(for operation IDs) bit in the DiscriminantID or
OperationID struct when the object key or operation ID is first
sent. Each side must then assign the next available index to that object
or operation. The space of operations is separate from the space of object
ids, so that a total of 16383 possible values is available for memoizing
of discriminant objects, and 16383 different possible values for memoizing
of operations.
Note that the index is passed implicitly, so both sides of the connection must synchronize their use of indices.
A shared set of indices may be loaded into the connection by some mechanism before any messages are sent. This specification does not define a mechanism for doing so.
Reply Message
Reply header (pseudo-C):
typedef struct {
Boolean control_msg : 1; /* == FALSE */
Boolean ext_hdr_present : 1; /* True if ext hdr list present */
ReplyStatus : 2;
Unused reply_1 : 4;
unsigned serial_no : 24; /* serial # from Request */
} ReplyMsgHeader; /* 4 bytes total */
The actual message consists of the following fields:
[ ReplyMsgHeader ]
[ extension header list, if any ]
[ exception ID (32-bit unsigned), if any ]
[ explicit output parameter values, if any, padded to 4 byte boundary ]
InitializeConnection Message
InitializeConnection header (pseudo-C):
typedef struct {
Boolean control_msg : 1; /* == TRUE */
ControlMsgType msg_type : 3; /* == InitializeConnection */
Unused verify_1 : 4;
ProtocolVersion version : 8; /* what version of the protocol? */
unsigned server_id_len : 16; /* length of server ID */
} InitializeConnectionMsgHeader;
The actual message consists of the following fields:
[ InitializeConnectionMsgHeader ]
[ server_id_len-length server ID for supposed callee, padded
to 4-byte boundary ]
This message is sent from caller to callee as the first message of the
connection. It is used to pass the server ID of the connection from client
to server, so that both sides understand what the omitted prefix portion
of discriminant IDs is. If the server ID received by the callee is not the
correct server ID for the callee (i.e., the callee has objects which do not
have that prefix in their object IDs), the callee should terminate the
connection, with the appropriate reason. The server ID is passed as an XDR
fixed-length opaque data value of the length specified in
server_id_len.
TerminateConnection Message
TerminateConnection header (pseudo-C):
typedef struct {
Boolean control_msg : 1; /* == TRUE */
ControlMsgType msg_type : 3; /* == TerminateConnection */
TerminationCause cause: 4; /* why connection terminated */
unsigned serial_no : 24; /* last request processed/sent */
} TerminateConnectionMsgHeader;
The actual message consists simply of the header; it provides for graceful connection shutdown. It is sent either from the caller to the callee, or from the callee to the caller, and informs the other party that it is cancelling the connection, for one of these reasons:
InitializeConnection message with the wrong major version for
the protocol;
InitializeConnection message with the
wrong server ID;
The serial_no field contains the serial number of the last message
completely processed by the caller (when TerminateConnection
is sent from caller to callee), or the serial number of the last message
sent by the callee (when sent from callee to caller). No further messages
should be sent on the connection by a sender of a
TerminateConnection message after it has been sent, or by a
receiver of TerminateConnection messsage after it has been received.
DefaultCharset Message
DefaultCharset header (pseudo-C):
typedef struct {
Boolean control_msg : 1; /* == TRUE */
ControlMsgType msg_type : 3; /* == DefaultCharset */
Unused bits_12: 12; /* unused */
unsigned charset_mibenum : 16; /* default charset */
} DefaultCharsetMsgHeader;
This message is sent by either side of a connection to establish a default
charset for subsequent messages sent by that side of the connection. The
charset defines how string values are marshalled as octet sequences. The
default charset defines the default marshalling, unless overridden by an
explicit charset in a string value. Each side of the connection may establish
a default charset independently of the other side of the connection; the
default charset only applies to string values in messages coming from that
side. A new value of the default charset may be established at any time by
sending another DefaultCharset message.
The data value format used for parameters is the XDR format specified in Internet RFC 1832. However, we extend the XDR specification with one additional type, called flagged variable-length opaque data. It is similar to XDR's regular variable-length opaque data, except that the high-order bit of the length field is used as a flag bit, instead of being part of the length. This means that flagged variable-length opaque data can only carry opaque data of lengths less than or equal to (2**31)-1.
0 1 2 3 4 5 ...
++----+-----+-----+-----+-----+-----+...+-----+-----+...+-----+
flag -->|| length n |byte0|byte1|...| n-1 | 0 |...| 0 |
bit ++----+-----+-----+-----+-----+-----+...+-----+-----+...+-----+
||<------31 bits------->|<------n bytes------>|<---r bytes--->|
|<----n+r (where (n+r) mod 4 = 0)---->|
FLAGGED VARIABLE-LENGTH OPAQUE
Values of type BOOLEAN are passed as XDR bool.
Values of enumeration types are passed as XDR enum. Each enumeration
value is assigned its ordinal value as it appears in the declaration of the
enumeration type, starting with the value `one'.
Values of fixed-point types are passed by passing the value of the numerator. We define a number of special cases for efficient marshalling of common integer types, as well as a general case for passing values of fixed-point types that are not covered by the special cases.
Special cases:
integer.
unsigned integer.
hyper
integer.
unsigned hyper
integer.
General case:
The numerator of the value is passed as XDR flagged variable-length
opaque data, with the bytes of the data containing the value expressed
as a base-256 number, in big-endian order; that is, with the most significant
digit of the value first. The flag bit is used to carry the sign; the flag
bit is 0 for a positive number or zero, and 1 for a negative number.
We define a number of special cases for efficient marshalling of common floating-point types, as well as a general case for passing values of floating-point types that are not covered by the special cases.
Special cases:
floating-point.
double-precision floating-point.
fixed-length opaque data, containing the floating-point value
in the format specified in the UNIX System V Application Binary Interface
Intel 386 Processor Supplement (Intel ABI) document: the 63 bits of the fraction
occupy the first 7 bytes in little-endian order plus the low seven bits of
the eighth byte; the 1 bit explicit leading significand bit occupies the
high-order bit of the eighth byte; the 15 bits of the exponent occupy the
ninth byte and the low-order bits of the tenth byte, in little-endian order;
the sign bit occupies the high-order bit of the tenth byte; the eleventh
and twelfth bytes are unused, and should contain zero values.
quadruple-precision floating-point.
General case:
Values of floating-point types not matching the special cases identified
above are passed as a value of the XDR struct type
GeneralFloatingPointValue, which has the following definition:
/* XDR */
enum { Normal = 1, NotANumber = 2, Infinity = 3 } FloatingPointValueType;
struct {
flagged opaque FixedPointSignAndSignificand<>;
flagged opaque FixedPointExponent<>;
} NormalFloatingPointValue;
union switch (FloatingPointValueType disc) {
case Normal: NormalFloatingPointValue value;
case NotANumber: void;
case Infinity: void;
} GeneralFloatingPointValue;
The two fields of the NormalFloatingPointValue struct each contain
an on-the-wire representation of a fixed-point value of the fixed-point type
(denominator=1, no-mininum-numerator, no-maximum-numerator). The
FixedPointSignAndSignificand field contains the sign of the
floating-point value as the sign, and the actual significand as the absolute
value of the fixed-point value. The FixedPointExponent field
contains the exponent of the floating-point value.
Each string value sent in this protocol has a charset [RFC 2278]
associated with it, identified by the charset's IANA-assigned MIBEnum value.
Each side of a session may establish a default charset by sending
the DefaultCharset message. String values that use the default
character set do not contain explicit charset information; string values
that use a charset other than the default charset contain the MIBEnum value
for the charset, along with the bytes of the string.
We send a string value as a value of XDR flagged variable-length opaque
data. If the flag bit is 1, the first two bytes of the string value
are the MIBEnum of the charset, high-order byte first; the remaining bytes
are the bytes of the string. If the flag bit is 0, the bytes of opaque data
simply contain the bytes of the string; the charset is the default charset
for the session. It is a marshalling error to send a string value with a
flag bit of 0 over a session for which no default charset has been established.
To avoid context-sensitivity in marshalling a string, it is always valid
to marshal a string with an explicit charset value, even if the charset value
is the same as the default charset for the session. When marshalling a string
into a pickle, the charset should always be explicitly included.
Values of sequence types are passed as XDR variable-length arrays,
with one exception: Sequences of any fixed-point type with a minimum numerator
greater than or equal to 0, and a maximum numerator less than or equal to
255, are passed as XDR variable-length opaque data, with one
numerator value per octet.
Values of array types are passed as XDR fixed-length arrays,
with one exception: Arrays of any fixed-point type with a minimum numerator
greater than or equal to 0, and a maximum numerator less than or equal to
255, are passed as XDR fixed-length opaque data, with one numerator
value per octet. Values of array types are passed as XDR fixed-length
arrays, with one exception:
Values of record types are passed as XDR struct.
Values of union types are passed as XDR union, with the union
discriminant being the zero-based ordinal value for the encapsulated value's
type.
A pickle is passed as an XDR variable-length opaque data, containing
the type ID of the pickled value's type, followed by the XDR-marshalled pickled
value. To save pickle space for common value types used in metadata, we define
a packed format for the type ID marshalling. A type ID is marshalled into
a pickle as a 32-bit header, in an XDR unsigned integer, possibly
followed by an XDR fixed-length opaque data, containing the
string form of the type ID of the pickled type. The header has the following
internal structure:
/* Pseudo-C */
typedef struct {
unsigned version : 8;
PickleTypeKind type_kind : 8;
unsigned type_id_len : 16;
} TypeIDHeader;
The version field gives the version number of the pickle format;
the type_kind field contains a value from the enum
/* Pseudo-C */
typedef enum {
TypeKind_unconstrained = 0, /* anything not covered by other type kinds... */
TypeKind_boolean = 1, /* BOOLEAN */
TypeKind_s8 = 2, /* FIXED-POINT DENOM=1 MIN-NUM=-128 MAX-NUM=127 */
TypeKind_s16 = 3, /* FIXED-POINT DENOM=1 MIN-NUM=-32768 MAX-NUM=32767 */
TypeKind_s32 = 4, /* FIXED-POINT DENOM=1 MIN-NUM=-2147483648 MAX-NUM=2147483647 */
TypeKind_s64 = 5, /* FIXED-POINT DENOM=1 MIN-NUM=-9223372036854775808
MAX-NUM=9223372036854775807 */
TypeKind_u8 = 6, /* FIXED-POINT DENOM=1 MIN-NUM=0 MAX-NUM=255 */
TypeKind_u16 = 7, /* FIXED-POINT DENOM=1 MIN-NUM=0 MAX-NUM=65535 */
TypeKind_u32 = 8, /* FIXED-POINT DENOM=1 MIN-NUM=0 MAX-NUM=4294967296 */
TypeKind_u64 = 9, /* FIXED-POINT DENOM=1 MIN-NUM=0 MAX-NUM=18446744073709551616 */
TypeKind_ieee_float32 = 10, /* FLOATING-POINT SIGNIFICAND-SIZE=24 EXPONENT-BASE=2
MAXIMUM-EXPONENT-VALUE=127 MINIMUM-EXPONENT-VALUE=-126
HAS-NOT-A-NUMBER=TRUE HAS-INFINITY=TRUE
DENORMALIZED-VALUE-ALLOWED=TRUE HAS-SIGNED-ZERO=TRUE */
TypeKind_ieee_float64 = 11, /* FLOATING-POINT SIGNIFICAND-SIZE=53 EXPONENT-BASE=2
MAXIMUM-EXPONENT-VALUE=1023 MINIMUM-EXPONENT-VALUE=-1022,
HAS-NOT-A-NUMBER=TRUE HAS-INFINITY=TRUE
DENORMALIZED-VALUE-ALLOWED=TRUE HAS-SIGNED-ZERO=TRUE */
TypeKind_i_default_str = 12, /* STRING LANGUAGE="i-default" */
TypeKind_object = 13, /* local or remote object */
...
/* other types like Date, etc, should be added here... */
...
} PickleTypeKind;
If the value of type_kind is
TypeKind_unconstrained, the value of type_kind_len
is the length of a value of XDR type fixed-length opaque data,
containing the full string type ID of the type, which immediately follows
the header. Otherwise, no opaque data is marshalled.
For the purposes of marshalling, pickles have no default charset; this means that strings marshalled into a pickle should always contain an explicit charset. Pickles should be considered a single "message" for the purposes of marshalling aliased reference types.
Optional types are passed as XDR optional-data.
The scope of aliasing in this protocol is the message, as in Java RMI, rather than the call, as in DCE RPC. That is, aliasing occurs only within the context of a single invocation or result, rather than across a full invocation-result pair. For the purposes of marshalling, a pickle scope should be considered a single message scope.
Each unique value of an aliased type that is marshalled is assigned a 32-bit
unsigned integer value, unique in the scope of aliasing, called its
aliased identifier. This identifier is marshalled as an XDR
unsigned integer. If the aliased value has not previously been
sent in this scope, its value is then marshalled as a value of its base type
would be. Note that this means that the full value of every aliased type
is sent only once in a scope; subsequent occurrences send only the aliased
identifier.
[ XXX - how to handle overflow of aliased value cache? ]
An instance of an object type is passed as the state of the object type, which also contains information about the actual type of the value. For remote object types, this state is followed by the object identifier, and optionally information about how the instance may be contacted.
When marshalling the state of an object, it's important to distinguish two important types of the value: the parameter type, which is the type that both sides of the session expect the value to have, and the actual type of the value, which is the most-derived type of the object, and may be a subtype of the parameter type. If the actual type is different from the parameter type, extra information must be passed along with the value to allow the receiver to properly distinguish the type and its associated data. However, if the actual type is the same as the parameter type, some of this information can be omitted.
We pass the state of the object type as the type ID of the most-derived-type
of the object, followed by the state attributes of each type of the object.
The type ID is passed as one of three values, depending on the following
conditions:
void.
variable-length
opaque data.
variable-length
opaque data containing the type ID.
The state attributes are marshalled in one of two ways:
structure with the
attributes as the components of the structure. The value of each attribute
is marshalled directly according to the type of the attribute.
structure values, each containing the state for one of the types
of the instance. Types of the instance which have no associated state do
not appear in this sequence. An XDR expression of the sequence would be the
following:
/* XDR */
struct {
opaque type_id<0xFFFF>;
opaque state<>;
} TypeState;
typedef TypeState StateSequence<>;
The type_id field contains the type ID for that type of the the object
value. The variable-length opaque data field state contains the values
of the attributes of the state marshalled as an XDR structure,
where the components of the structure are the attributes of the state.
In the case of a remote object type, the server ID, instance handle and contact
info for the value are passed as a value of the following XDR structure type
RemoteObjectInfo:
/* XDR */
typedef string ContactInfo<0xFFFF>;
struct {
opaque server_id<>;
opaque instance_handle<>;
ContactInfo cinfos<>;
} RemoteObjectInfo;
where server_id is a identifier for the server which supports the desired object, and instance_handle is a server-relative name for the object. The cinfos field contains zero or more pieces of information about the way in which the object needs to be contacted, including information such as whether various transport layers are involved.
UnknownProblem
Exception Code: 0
ISL Values: None
ImplementationLimit
Exception Code: 1
ISL Values: None
The request could not be properly addressed because of some implementation resource limit on the callee side.
SwitchConnectionCinfo
Exception Code: 2
ISL Values: NEW-CINFO : HTTP-ng-w3ng.CinfoString
This exception requests the caller to upgrade the connection protocol and
transport information to the cinfo specified as the argument, and re-try
the call. This is the equivalent of the UPGRADE message in HTTP
1.1, and the RELOCATE_REPLY message in CORBA GIOP.
Marshal
Exception Code: 3
ISL Values: None
A marshalling problem was encountered.
NoSuchObjectType
Exception Code: 4
ISL Values: None
The object type of the operation was unknown at the server.
NoSuchMethod
Exception Code: 5
ISL Values: None
The object type of the operation was known at the server, but did not contain the indicated method.
NoSuchObject
Exception Code: 6
ISL Values: None
The specified discriminant object was not available at the server.
InvalidType
Exception Code: 7
ISL Values: None
The object specified by the discriminant did not participate in the type specified in the operation.
Rejected
Exception Code: 8
ISL Values: REASON : OPTIONAL SimpleString
The server refused to process the request. It may return a string giving a reason for the rejection.
OperationOrDiscriminantCacheOverflow
Exception Code: 9
ISL Values: None
The request caused the receiver's cache of operations or discriminants to overflow. The sender may retry the request with uncached operation and discriminant values; subsequent requests should not cache any additional operation or discriminant values, but may continue to use previously successfully cached values.
Does this protocol need to assign serial numbers to requests and replies?
We do so in order to be able to cancel operations by serial number, and to
be able to return reply messages out of order. The first problem, that of
cancelling operations, could be dealt with by keeping track of serial numbers
implicitly, and using an explicit serial number only in the
CancelRequest message. Doing this would imply that the replies
would have to be returned in the order in which the requests were passed,
but would allow us to have 6 byte request messages (4 bytes if we count the
discriminant as part of the arguments, instead of part of the header), and
4 byte reply messages. Thus the only real purpose for serial numbers is to
allow replies to be returned out of order (and possibly to make debugging
the protocol easier). There are other deeper unanswered questions here about
the serialization semantics of the protocol. For instance, should the callee
wait until dispatching a reply to one request until beginning to process
the next one?
The current answer to these questions is that it is highly useful to allow a threaded callee to process multiple requests in parallel, and to allow it to return requests out of order. Thus serial numbers are useful. We assume that higher-level protocols desiring serialization will provide a serialization context as part of the context of the call, and that serialization will be handled at either a higher or lower level.
A great deal of the traffic over this protocol may consist of values of type PICKLE (the equivalent of object-by-value, or of HTTP's MIME-encapsulated body type) or of some object type. It is tempting to introduce a form of memoizing for these value types, similar to that used for request discriminants. There are two reasons not to do so:
struct {
boolean use_cached_value : 1;
boolean cache_this_value : 1;
union {
unsigned int url_len : 30;
unsigned int cache_key : 30;
} v;
};
either by itself (if use_cached_value is set), or followed by
an XDR fixed length opaque value containing the URL for the object (if
use_cached_value is not set). This type of variable structure
has no equivalent in XDR. On the other hand, it could well be argued that
since we are marshalling an object type, something not explicitly covered
by XDR, that we are simply providing an extension to XDR, in the spirit of
the marshalling. We could even use a simpler construct, such as XDR union.
TerminateConnection.
Neither of these arguments seems overwhelmingly powerful.
Open issues:
http-ng://foo.bar.com/object-id might be used to
indicate that object object-id is available at
foo.bar.com, and that the client should use some protocol
negotiation protocol to work out the exact shape of the cinfo.
Proposed: URLs for HTTP-ng objects will be of the form
w3ng:SERVER-ID/INSTANCE-HANDLE[;type=TYPE][;cinfo=CINFO]
where SERVER-ID is a identifier for the server which supports the desired object; INSTANCE-HANDLE is a server-relative name for the object; TYPE is the type ID for the most derived type of the object; and CINFO is information about the way in which the object needs to be contacted, including information such as whether various transport layers are involved. This form has the virtue of becoming a URN if the optional CINFO and TYPE fields are omitted.
The syntax of cinfo currently follows the ILU definition. Each cinfo string has the form described below (where brackets indicate optionality, an <ALPHANUMERIC-ID> is an identifier composed of ASCII lowercase alphabetic and numeric characters, beginning with a lowercase alphabetic character, and a <NON-UNDERSCORE-STRING> is any string of ASCII characters not containing the underscore character '_'):
<cinfo> := <pinfo> '@' <tinfo-stack> <pinfo> := <scheme> [ '_' <parms> ] <scheme> := <ALPHANUMERIC-ID> <parms> := <parm> [ '_' <parms> ] <parm> := <NON-UNDERSCORE-STRING> <tinfo-stack> := <tinfo> [ '=' <tinfo-stack> ] <tinfo> := <scheme> [ '_' <parms> ]
w3ng Pinfo
The current syntax of the pinfo string for the ILU implementation of the
w3ng wire protocol is
<scheme> := 'w3ng' <parms> := <major-version> [ '.' <minor-version> ]
where <major-version> and
<minor-version> are numbers between 0 and 15. If the
<minor-version> is not specified, it defaults to 0.
w3mux Tinfo
The current syntax of the tinfo string for the ILU implementation of the
w3mux transport layer is
<scheme> := 'w3mux' <parms> := <channel> '_' <endpoint>
where <channel> is a protocol ID number [MUX], and
<endpoint> is a UUID string for an endpoint. The size
of the <endpoint> string must be less than 1000 bytes.
tcp Tinfo
The current syntax of the tinfo string for the ILU implementation of the
tcp transport layer is
<scheme> := 'tcp' <parms> := <host> '_' <port>
where <host> is string of less than 1000 bytes indicating
the IP address or hostname of the remote machine, and
<port> is the TCP port on which the host is listening.
sunrpcrm Tinfo
The current syntax of the tinfo string for the ILU implementation of the
sunrpcrm transport layer is
<scheme> := 'sunrpcrm'
No parameters are defined. This layer implements the ONC RPC record-marking scheme on top of a reliable byte stream, as defined in section 10 of the ONC RPC RFC [ONC RPC].
RFC 2278: http://info.internet.isi.edu:80/in-notes/rfc/files/rfc2278.txt
XDR [RFC 1832]: http://info.internet.isi.edu:80/in-notes/rfc/files/rfc1832.txt
ONC RPC [RFC 1831]: http://info.internet.isi.edu:80/in-notes/rfc/files/rfc1831.txt
ISL: ftp://ftp.parc.xerox.com/pub/ilu/2.0a12/manual-html/manual_2.html
WD-HTTP-NG-arch-model (work in progress): http://www.w3.org/TR/1998/WD-HTTP-NG-architecture
MUX (work in progress): http://www.w3.org/TR/1998/WD-mux
ILU: ftp://ftp.parc.xerox.com/pub/ilu/2.0a12/manual-html/manual_2.html
Bill Janssen
Xerox Palo Alto Research Center
3333 Coyote Hill Rd
Palo Alto, CA 94304
Phone: (650) 812-4763
FAX: (650) 812-4777
Email: janssen@parc.xerox.com
HTTP: http://www.parc.xerox.com/istl/members/janssen/
BOOLEAN
Boolean (pseudo-C enum
type)
ControlMsgType (pseudo-C
enum type)
DefaultCharsetMsgHeader (pseudo-C
struct type)
DiscriminantID (pseudo-C
struct type)
FloatingPointValueType (XDR
type)
GeneralFloatingPointValue
(XDR type)
HTTP-ng-w3ng.ExtensionHeader
(ISL type)
HTTP-ng-w3ng.ExtensionHeaderList
(ISL type)
HTTP-ng-w3ng.SimpleString
(ISL type)
ImplementationLimit (system
exception)
InitializeConnection
message
InitializeConnectionMsgHeader
(pseudo-C struct type)
InvalidType (system
exception)
Marshal (system exception)
NormalFloatingPointValue
(XDR type)
NoSuchMethod (system
exception)
NoSuchObject (system
exception)
NoSuchObjectType (system
exception)
OperationID (pseudo-C struct
type)
ProtocolVersion (pseudo-C
struct type)
Rejected (system exception)
Reply message
ReplyMsgHeader (pseudo-C
struct type)
ReplyStatus (pseudo-C enum
type)
Request message
RequestMsgHeader (pseudo-C
struct type)
StateSequence (XDR type)
Success subtype of Reply
SwitchConnectionCinfo (system
exception)
SystemException subtype of
Reply
TerminateConnection message
TerminateConnectionMsgHeader
(pseudo-C struct type)
TerminationCause (pseudo-C
enum type)
TypeState (XDR type)
UnknownProblem (system
exception)
Unused (pseudo-C alias
type)
UserException subtype of
Reply
w3ng URL form
RemoteObjectInfo