RIFF

A RIFF file is a tagged file format. It has a specific container format (a chunk) with a header that includes a four-character tag (FourCC) and the size (number of bytes) of the chunk. The tag specifies how the data within the chunk should be interpreted, and there are several standard FourCC tags. Tags consisting of all capital letters are reserved tags. The outermost chunk of a RIFF file has a RIFF tag; the first four bytes of chunk data are an additional FourCC tag that specify the form type and are followed by a sequence of subchunks. In the case of a WAV file, the additional tag is WAVE. The remainder of the RIFF data is a sequence of chunks describing the audio information.

The advantage of a tagged file format is that the format can be extended later while maintaining backward compatibility.^[11] The rule for a RIFF (or WAV) reader is that it should ignore any tagged chunk that it does not recognize.^[12] The reader will not be able to use the new information, but the reader should not be confused.

The specification for RIFF files includes the definition of an INFO chunk. The chunk may include information such as the title of the work, the author, the creation date, and copyright information. Although the INFO chunk was defined for RIFF in version 1.0, the chunk was not referenced in the formal specification of a WAV file. Many readers had trouble processing this. Consequently, the safest thing to do from an interchange standpoint was to omit the INFO chunk and other extensions and send a lowest-common-denominator file. There are other INFO chunk placement problems.

RIFF files were expected to be used in international environments, so there is CSET chunk to specify the country code, language, dialect, and code page for the strings in a RIFF file.^[13] For example, specifying an appropriate CSET chunk should allow the strings in an INFO chunk (and other chunks throughout the RIFF file) to be interpreted as Cyrillic or Japanese characters.

RIFF also defines a JUNK chunk whose contents are uninteresting.^[14] The chunk allows a chunk to be deleted by just changing its FourCC. The chunk could also be used to reserve some space for future edits so the file could be modified without being resized. A later definition of RIFF introduced a similar PAD chunk.^[15]

RIFF WAVE

The top-level definition of a WAV file is:^[16]

<WAVE-form> → RIFF('WAVE'
                   <fmt-ck>            // Format of the file
                   [<fact-ck>]         // Fact chunk
                   [<cue-ck>]          // Cue points
                   [<playlist-ck>]     // Playlist
                   [<assoc-data-list>] // Associated data list
                   <wave-data> )       // Wave data

The top-level RIFF form uses a WAVE tag. It is followed by a mandatory <fmt-ck> chunk that describes the format of the sample data that follows. This chunk includes information such as the sample encoding, number of bits per channel, the number of channels, and the sample rate.

The WAV specification includes some optional features. The optional <fact-ck> chunk reports the number of samples for some compressed coding schemes. The <cue-ck> chunk identifies some significant sample numbers in the wave file. The <playlist-ck> chunk allows the samples to be played out of order or repeated rather than just from beginning to end. The associated data list (<assoc-data-list>) allows labels and notes to be attached to cue points; text annotation may be given for a group of samples (e.g., caption information).

Finally, the mandatory <wave-data> chunk contains the actual samples in the format previously specified.

Note that the WAV file definition does not show where an INFO chunk should be placed. It is also silent about the placement of a CSET chunk (which specifies the character set used).

The RIFF specification attempts to be a formal specification, but its formalism lacks the precision seen in other tagged formats. For example, the RIFF specification does not clearly distinguish between a set of subchunks and an ordered sequence of subchunks. The RIFF form chunk suggests it should be a sequence container. Sequencing information is specified in the RIFF form of a WAV file consistent with the formalism: "However, <fmt-ck> must always occur before <wave-data>, and both of these chunks are mandatory in a WAVE file."^[17] The specification suggests a LIST chunk is also a sequence: "A LIST chunk contains a list, or ordered sequence, of subchunks."^[18] However, the specification does not give a formal specification of the INFO chunk; an example INFO LIST chunk ignores the chunk sequence implied in the INFO description.^[19] The LIST chunk definition for <wave-data> does use the LIST chunk as a sequence container with good formal semantics.

The WAV specification supports, and most WAV files use, a single contiguous array of audio samples. The specification also supports discrete blocks of samples and silence that are played in order. The specification for the sample data contains apparent errors:^[20]

The <wave-data> contains the waveform data. It is defined as follows:
  <wave-data>  → { <data-ck> | <data-list> }
  <data-ck>    → data( <wave-data> )
  <wave-list>  → LIST( 'wavl' { <data-ck> | // Wave samples
                                <silence-ck> }... ) // Silence
  <silence-ck> → slnt( <dwSamples:DWORD> ) // Count of silent samples

Apparently <data-list> (undefined) and <wave-list> (defined but not referenced) should be identical. Even with this resolved, the productions then allow a <data-ck> to contain a recursive <wave-data> (which implies data interpretation problems). To avoid the recursion, the specification can be interpreted as:

<wave-data>  → { <data-ck> | <wave-list> }
  <data-ck>    → data( <bSampleData:BYTE> ... )
  <wave-list>  → LIST( 'wavl' { <data-ck> | // Wave samples
                                <silence-ck> }... ) // Silence
  <silence-ck> → slnt( <dwSamples:DWORD> ) // Count of silent samples

WAV files can contain embedded IFF "lists", which can contain several "sub-chunks".^[21][22][23]