HTTP/2.0 Discussion: Binary Optimized Header Encoding
draft-snell-httpbis-bohe-03

Abstract

This memo describes a proposed alternative encoding for headers within SPDY SYN_STREAM, SYN_REPLY and HEADERS frames.

Status of This Memo

This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."

This Internet-Draft will expire on August 05, 2013.

Copyright Notice

Copyright (c) 2013 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document.

Table of Contents

• 1. Binary Optimized Header Encoding
• 2. Value Encoding
• 3. Example Headers
• 4. Security Considerations
• 5. Normative References
• Author's Address

1. Binary Optimized Header Encoding

Binary Optimized Header Encoding is a proposed alternative serialization for headers within SPDY SYN_STREAM, SYN_REPLY and HEADERS frames that is designed to optimize generation, consumption and processing of the most commonly used HTTP headers.

Alternate Header Block Serialization:

 +-------+---------+
 |  NUM  | Headers |
 +-------+---------+

'NUM' is a single unsigned 8-bit value indicating the total number of distinct headers contained in the encoded block.

'Headers' consists of the complete set of encoded headers. The format for each header is determined by whether the header name (the key) and the header value are included in the static "default dictionary" defined by this specification. The first two bits of each encoded header record identify the specific header format used.

Registered Key + Registered Value Pair

  +--+---+--+
  |00|CNT|ID| 
  +--+---+--+

If the header name and value are both included in the default dictionary, the associated index of the header as provided by the dictionary is encoded as a variable-length, unsigned integer in lists of no more than 63 items. For example, the headers ":path=/", ":method=GET" and ":scheme=https" would be compactly encoded as: 03 03 0F 06

Registered Key + Unregistered Value

  +--+---+--+---+---+
  |01|FLG|ID|LEN|VAL|
  +--+---+--+---+---+

If the header name is included in the default dictionary, but the value given is not, the header is encoded using five fields: the two-bit opcode (01) followed by six flag bits as defined below, followed by the numerically lowest index of the named header as provided by the default dictionary encoded as a variable-length, unsigned integer, followed by another variable-length, unsigned integer specifying the length of the value, followed by the encoded value.

Unregistered Key + Value Pair

  +--+---+------+---+------+---+
  |10|FLG|LENkey|key|LENval|val|
  +--+---+------+---+------+---+

If the header name and value are not included in the default dictionary, the header is encoded using six distinct fields: the two-bit opcode (10) followed by six flag bits as defined below, followed by a variable-length, unsigned integer specifying the key length, followed by the ISO-8859-1 encoded key name, followed by a variable-length, unsigned integer specifying the value length, followed by the encoded value.

Flags:

  FLG = 000000
        ||||||
        |||||Reserved
        ||||Date (Numeric val represents a date)
        |||Numeric (val is one or more uvarints)
        ||Huffman-encoded
        |UTF-8 or ISO8859-1 (1 or 0)
        Binary | Text (1 or 0)

When the opcode is either 01 or 10, the six-bit FLG field is used to specify additional properties about the header value. When the opcode is 00, the FLG bits are unused and MUST NOT be set.

Moving from the most significant bit to the least, the flags are:

• Binary or Text (0x32) - When set, the value is to be interpreted as binary data. When unset, the value is to be interpreted as encoded-text.
• UTF-8 or ISO-8859-1 (0x16) - When set, indicates that the text-encoded value is UTF-8 encoded. Otherwise, ISO-8859-1 is assumed. Only used if the Binary bit (0x32) is unset.
• Huffman-encoded (0x08) - When set, indicates that the text-encoded value has been static huffman encoded. The static dictionary used for the encoding is dependent on whether the text is UTF-8 or ISO-8859-1 encoded. Only used if the Binary bit (0x32) is unset.
• Numeric (0x04) - When set, indicates that the value is encoded as a variable-length unsigned integer. Only used if the Binary bit (0x32) is set (indicating a binary value).
• Date (0x02) - When set, indicates that the value is a date and time encoded as a variable-length unsigned integer representing the total number of seconds that have ellapsed since midnight on January 1st, 1990, GMT. Only used if the Binary (0x32) and Numeric (0x04) bits are both set.
• The final bit (0x01) is reserved.

2. Value Encoding

Header values can be encoded as one of four distinct types: Binary, Text, Numeric or Date. Each is indicated by the FLG field.

• Binary Header values are encoded as a raw, uninterpreted sequence of binary octets.
• Numeric values are encoded as variable-length, unsigned integers.
• Date values are encoded as variable-length, unsigned integers representing the total number of seconds that have ellapsed since midnight on January 1st, 1990.
• Text values are encoded as a sequence one or more length-prefixed octet sequences representing ISO-8859-1 or UTF-8 encoded characters, conforming to the following structure:

  string_val = COUNT 1*(LENGTH VAL)
      

Where COUNT is an unsigned 8-bit value specifying the number of octet sequences, LENGTH is an unsigned variable-length integer specifying the length of each sequence, and VAL is the specific sequence of encoded octets.

3. Example Headers

Assuming the following (incomplete) default dictionary:

The header :version=2.0 would be encoded as:

  01 1D |..|

The header :method=GET would be encoded as:

  01 06 |..|

The headers :version=2.0 and :method=GET would be encoded as:

  02 06 1D |..|

The header :method=foo would be encoded as:

  56 05 05 01 03 66 6F 6F |.....foo|

The header foo=bar would be encoded as:

  b6 03 66 6f 6f 05 01    |..foo...|
  03 62 61 72             |.bar|

4. Security Considerations

TBD

5. Normative References

Author's Address