Help language development. Donate to The Perl Foundation

HTTP::HPACK zef:jnthn last updated on 2023-04-12

736d745582e3aa84bebefba2462151fcee686d44/

HTTP::HPACK

Build Status

Synopsis

Decoding:

my $decoder = HTTP::HPACK::Decoder.new;
my @headers = $decoder.decode-headers($buf);
say "{.name}: {.value} ({.indexing})" for @headers;

Encoding:

my @headers = 
    HTTP::HPACK::Header.new(
        name  => ':method',
        value => 'GET'
    ),
    HTTP::HPACK::Header.new(
        name     => 'password',
        value    => 'correcthorsebatterystaple',
        indexing => HTTP::HPACK::Indexing::NeverIndexed
    );
my $encoder = HTTP::HPACK::Encoder.new;
my $buf = $encoder.encode-headers(@headers);

Description

HPACK is the HTTP/2 header compression algorithm. This module implements encoding (compression) and decoding (decompression) of the HPACK format, as specified in RFC 7541. A HTTP/2 connection will typically have an instance of the decoder (to decompress incoming headers) and an instance of the encoder (to compress outgoing headers).

Notes on specific features

Huffman compression

Decoding of headers compressed using the Huffman codes (set out in the RFC) takes place automatically. By default, the encoder will not apply Huffman compression. To enable this, construct it with the huffman option set to True:

my $encoder = HTTP::HPACK::Encoder.new(:huffman);

Dynamic table management

The dynamic table size can be limited by passing the dynamic-table-limit option when constructing either the encoder or decoder:

my $decoder = HTTP::HPACK::Decoder.new(dynamic-table-limit => 256);

It is also possible to introspect the current dynamic table size:

say $decoder.dynamic-table-size;

The size is computed according to the algorithm in RFC 7541 Section 4.1.

Thread safety

Instances of HTTP::HPACK::Header are immutable and so safe to share and access concurrently. Instances of HTTP::HPACK::Decoder and HTTP::HPACK::Encoder are stateful (as a result of the dynamic table), and so a given instance may not be used concurrently. This is not a practical problem, since headers may only be processed in the order they are being received or transmitted anyway.

Known issues