For ODF-Next we might take a look at whether we can make some simple structural changes that will allow applications to load and save documents faster. Options to consider include looking at incremental loads and saves.
Also, are there things we can do to better optimize very large spreadsheets?
It seems that in some cases there is a trade-off between open performance and save performance. For example, having one XML file per slide in a presentation would make a complete load faster, but would allow a faster save in cases where only some slides have changed. However, the dynamic of this may change with the prevalence. of multi-core processors on the desktop, where parsing two XML files might be done faster than parsing one large XML file.
We should ask implementors what bottlenecks they've observed in their applications when loading/saving ODF documents and see how they might be addressed by format enhancements.