One of the most important design decisions for MusicXML was choosing its scope. The goal was to represent as much music as possible that fell within a common tradition, making it neither too narrow nor too broad. The decision was to represent common Western music notation from the 17th century onwards. This includes present-day popular music notation and tablature, but excludes medieval music and music from other traditions and geographies, such as the many rich musical cultures in Asia and Africa.
Another aspect of the scope decision was the type of applications that would use the format. We wanted MusicXML to be used by any applications that would use a symbolic music format – not just notation editing, but music scanning, sequencing, performing, teaching, and musicological analysis. This had important implications for structure decisions as we will describe below.
Setting scope appropriately is essential to having an interchange language represent documents in a way that is faithful to the documents as understood by the people who created them. Western medieval music, for instance, has very different core concepts than today’s Western music. Pitch and rhythm are specified very differently, not only with different symbols, but with symbols that convey different meaning to the musicians that perform from them. For example, the barlines that are central to most Western music are absent from most medieval music, reflecting very different roles for meter. As Dumitrescu (2001) argues, the best approach is to represent medieval music in a separate language that is faithful to the original music, and then use exchange tools to allow conversion back and forth with modernized notation. Dumitrescu’s Corpus Mensurablis Musice Electronicum project is an example of such an XML language for representing the mensural notation of European music in the 14th through 16th centuries.
Differences with non-Western music are even greater. Different musical communities have different understanding of pitch and rhythm, and what is “the same” versus what is “different” between two musical performances. Statistical learning forms a large component of how we perceive and respond to music (Huron, 2006). The very different acoustic properties of the world’s musics suggest that you do not get universal understanding of music without similarly universal listening.
For symbolic music, the SMDL format is a canonical example of the dangers of setting scope too broadly. Its goals included trying to represent all music – both sound and “notation in any graphical form”. But no musician alive understands all music from all time periods. The resulting format is overly abstract, disempowering musicians at the expense of computer theorists. Most symbolic music software application developers are musicians, so making a format that musicians cannot understand alienates both the developer and user communities.