english  |  russian 
ogg mapping
This page specifies the way in which compressed FLAC data is encapsulated in an Ogg transport layer. It assumes basic knowledge of the FLAC format and Ogg structure and framing.

The original FLAC format includes a very thin transport system. This system of compressed FLAC audio data mixed with a thin transport has come to be known as 'native FLAC'. The transport consists of audio frame headers and footers which contain synchronization patterns, timecodes, and checksums (but notably not frame lengths), and a metadata system. It is very lightweight and does not support more elaborate transport mechanisms such as multiple logical streams, but it has served its purpose well.

The native FLAC transport is not a transport "layer" in the way of standard codec design because it cannot be entirely separated from the payload. Though the metadata system can be separated, the frame header includes both data that belongs in the transport (sync pattern, timecode, checksum) and data that belongs in the compressed packets (audio parameters like channel assignments, sample rate, etc).

This presents a problem when trying to encapsulate FLAC in other true transport layers; the choice has to be made between redundancy and complexity. In pursuit of correctness, a mapping could be created that removed from native FLAC the transport data, and merged the remaining frame header information into the audio packets. The disadvantage is that current native FLAC decoder software could not be used to decode because of the tight coupling with the transport. Either a separate decoding implementation would have to be created and maintained, or an Ogg FLAC decoder would have to synthesize native FLAC frames from Ogg FLAC packets and feed them to a native FLAC decoder.

The alternative is to treat native FLAC frames as Ogg packets and accept the transport redundancy. It turns out that this is not much of a penalty; a maximum of 12 bytes per frame will be wasted. Given the common case of stereo CD audio encoded with a blocksize of 4608 samples, a compressed frame will be 4-16 Kbytes. The redundancy amounts to a fraction of a percent.

In the interest of simplicity and expediency, the second method was chosen for the first official FLAC->Ogg mapping. A mapping version is included in the first packet so that a less redundant mapping can be defined in the future.

It should also be noted that support for encapsulating FLAC in Ogg has been present in the FLAC tools since version 1.0.1. However, the mappings used were never formalized and have insurmountable problems. For that reason, Ogg FLAC streams created with flac versions before 1.1.1 should be decoded and re-encoded with flac 1.1.1 or later (flac 1.1.1 can decode all previous Ogg FLAC files, but files made prior to 1.1.0 don't support seeking). Since the support for Ogg FLAC before FLAC 1.1.1 was limited, we hope this will not result in too much inconvenience.

Version 1.0 of the FLAC-to-Ogg mapping then is a simple identifying header followed by pure native FLAC data, as follows: It is intended that the first six bytes of any version of FLAC-to-Ogg mapping will share the same structure, namely, the four-byte signature and two-byte version number.

There is an implicit hint to the decoder in the mapping version number; mapping versions which share the same major version number should be decodable by decoders of the same major version number, e.g. a 1.x Ogg FLAC decoder should be able to decode any 1.y Ogg FLAC stream, even when x<y. If a mapping breaks this forward compatibility the major version number will be incremented.