Symbolic Formats Using XML
Around 2000, many people realized that XML technology could be a great way to finally create a successful interchange format for music notation and digital sheet music applications (Castan et al., 2001). The Recordare® MusicXML format was one of numerous proposals. Others included MML (Steyn, 2002), MEI (Roland, 2002), and WEDELMUSIC (Bellini and Nesi, 2001). Like SMDL before them, none of these contemporary XML proposals attracted support from commercial music applications.
In contrast, MusicXML has succeeded in becoming the de facto interchange standard for symbolic music formats. Figures 1 and 2 show the growth in MusicXML adoption between when MusicXML 0.5 was presented at XML 2001 (Good, 2001) and the status as of XML 2006. Even five years ago, MusicXML had achieved more success as a symbolic music interchange format than any prior effort besides MIDI:
Figure 1: MusicXML Adoption as of XML 2001
However, industry adoption is far more pervasive today:
Figure 2: MusicXML Adoption as of XML 2006
To see the difference that MusicXML makes to musicians, compare how a file transfers from Sibelius to Finale using MusicXML instead of MIDI. Sibelius and Finale are the two dominant professional music notation editors, and it is very common for users to want to exchange files between the two applications. A typical case is when a composer uses Sibelius, but a publisher, arranger, or engraver uses Finale.
Figure 3: Original Music as Entered in Sibelius 4.1
Figure 4: Music as Transferred to Finale 2007 via Standard MIDI File
Figure 5: Music as Transferred to Finale 2007 via MusicXML
Figure 3 shows the original Sibelius 4 file. Figure 4 shows the results of a MIDI transfer between Sibelius 4.1 and Finale 2007. Figure 5 shows the results of a MusicXML transfer between Sibelius 4.1 (exported using Recordare’s Dolet® 3.2 for Sibelius plug-in) and Finale 2007. The results in Figure 4 are so poor that the Finale user would likely need to start over and re-enter the notes manually. Not all transfers differ this dramatically, but MusicXML transfers between notation applications nearly always save considerable manual rework compared to MIDI transfers. This type of transfer is just one of many current uses of MusicXML within the world of commercial music applications (Good, 2006).
Why did MusicXML succeed when so many prior and contemporary efforts to create a standard interchange format for symbolic music files had failed? The entrenched music application vendors were initially no more receptive to the MusicXML effort than they had been to prior efforts – perhaps even less so, given the previous high-profile failures of both NIFF and SMDL. How was MusicXML able to support both established market leaders like Finale and newcomers like SharpEye Music Reader while still in beta? The answers may prove valuable for designers of potential interchange standards in other application domains, especially those with entrenched market leaders who may not see an interchange standard as being helpful for their business model.
I believe there are six primary lessons to draw from the MusicXML experience, which may be summarized as follows:
- Apply usability techniques to XML language design.
- Develop the format together with the software.
- Support a market leader early.
- Market to other software developers.
- Give format developers good support.
- Avoid overhead.
