Introduction
dcb is a content coding for HTTP: Brotli compression that uses a second file, a
dictionary, which the client already holds, so that a new version of a page or a script is
sent as little more than what changed. It is defined in RFC 9842, “Compression Dictionary Transport”,
together with its Zstandard counterpart dcz. A saved
response body is a .dcb file: a fixed header, then the compressed data. The file name is
not fixed; .dcb is the usual name for a saved body, which has no signature of its own
other than the header’s.
This is not Brotli with another name. A .br file is a
plain Brotli stream with no header at all; a .dcb file has a magic number and a hash, and its
stream is a shared Brotli stream (RFC 9841) that can copy from
the dictionary as well as from earlier output. The two have different headers, different tools make them,
and a .br reader cannot read one. That is why they are separate formats here: they do not
share a header, a signature or a decoder, and neither is a version of the other.
Structure
0Magic numberFF 44 43 42(0xFF, then “DCB”)4 bytes4SHA-256 hash of the dictionary the stream was made with32 bytes36Shared Brotli stream compressed with the dictionary as a raw prefix dictionarythe rest
The dictionary is a raw one: its bytes act as if they sat in front of the sliding window, so a
copy can reach back into them whatever the window size (RFC 9841 section 3.2). The stream’s window is at
most 16 MB, and a client that announces dcb must decode windows up to that. The file holds
only the dictionary’s hash, not the dictionary: a reader has to find the dictionary whose SHA-256 digest is
that hash, and the RFC says to check it before use. How the stream refers into the dictionary, and the
header and negotiation fields that go with it, are on the
DCB and DCZ reference page and the
shared dictionaries page.
In the explorer
Drop a .dcb file. The tree shows the header (magic and hash) and the stream, and the stream
opens as the Brotli viewer shows any stream: meta-blocks, prefix codes and commands.
To read the data, give the dictionary with it: drop both files at once (or a folder holding both). The
explorer hashes the files beside the .dcb file with SHA-256 and uses the one whose digest is
the header’s hash; the view says which file it was. Without the dictionary the stream is read as far as it
can be. Copies that reach past the window can only point into the dictionary, so each is listed (how far
back from the end of the dictionary it reaches, and how long it is), and the bytes before the first one
are decoded exactly. When the stream’s literal codes depend on the bytes before them, decoding has to stop
at the first such copy; the view says where and why.
Not described
Read against RFC 9842 and RFC 9841 in November 2026; the explorer was checked against the
brotli command-line tool, which compresses with a raw dictionary (-D).
Known, not yet covered
- Viewer: a dictionary of the other type RFC 9841 knows, a serialised dictionary with
custom word lists and transforms, is not what
dcbuses (RFC 9842 defines only raw dictionaries), so the viewer does not try one here. A serialised dictionary file has its own viewer and the stream decoder supports one. - Viewer: a dictionary larger than 128 MB is not read from the files given with the header.
- Reference: the HTTP side (
Use-As-Dictionary,Available-Dictionary, matching and caching) is summarised on the DCB and DCZ page but not set out rule by rule.
Links
Software
- brotli – Google’s command-line tool;
-Dcompresses against a raw dictionary, which gives the stream inside a dcb file (macOS, Linux, Windows) - Web browsers – ask for
dcbinAccept-Encodingwhen they hold a matching dictionary (Chromium-based browsers)