A great deal of effort has been expended on encoding musical data for notation, analysis, and query. More than 20 music codes are described in detail in Beyond MIDI: The Handbook of Musical Codes (Selfridge-Field, 1997); several dozen others exist. More than 1,000 complete works of classical music have been symbolically encoded for non-commercial use. For the full potential of music representation to be realized and music-information-retrieval procedures to be implemented, a common interchange standard is essential.
Existing methods of interchanging musical data range from formal schemes designed over many years but implemented only slightly to de facto procedures adapted to short-term needs. At the formal end of the range, Standard Music Description Language (SMDL) used principles of SGML to build a tagging system for elements of musical notation, musical performance, and “the logical core of the work,” in which many elements might duplicate those of the other musical phenomena but a few might be unique. Although the formal description was adopted by the ISO in 1996, implementations of SMDL have been few.
The NIFF file format (NIFF 1996; Grande 1997) was designed to allow the interchange of music-notation data between music-notation editing and publishing programs and music-scanning programs. Its intended range of applications ran from full music-publishing systems through simpler music-display systems to symbolic codes. Its original model was the SCORE parametric representation (Smith 1997). Although NIFF specifications were designed with industry support, neither of the original sponsors (Passport, Coda) implemented the format in commercial products.
At the informal end, MIDI is effectively the one interchange format in widespread use today. While sufficient for many performance applications, MIDI does not represent many concepts needed by musical applications concerned with printing and data retrieval. Its inadequacies reflect its origins as a hardware interface. It is inexact about enharmonic notation and too literal about performed note values, such that rhythmic values may deviate from the norms of visual grammar in images.
A frequent proposal has been to extend an existing code for use in the interchange of musical data between applications. In essence, MIDI has been that code. Its shortcomings for applications outside sound illustrate the pitfalls that could arise with any code developed for one purpose and “borrowed” for another. Musical information is enormously feature-rich. Codes work for a specific application when they represent all the variables relevant to that application. When the goal is changed, the variables relevant to the new application must magically be available.
Many existing music representations are inappropriate for interchange due to their narrow focus on one objective. They may be useful only as input codes (Selfridge-Field 1997: 571). Some would argue, conversely, that other representations, including SMDL, have attempted to represent music too broadly. SMDL has been unable to attract a large user group in part because it is difficult for potential users and tool-developers to see how SMDL might apply to their particular application. A standard that is defined generally enough to represent all music can only be made to work for a particular subset with great effort.
While XML seems to provide musicians with a representation scheme that holds the potential to eclipse both MIDI and formal interchange codes, such as NIFF and SMDL, as an interchange format, early experiments with XML have identified two areas in which some difficulties may lurk:
These procedural issues are both known from the world of music representation and data interchange prior to the advent of XML (see Hall 1997).