- Overall
- Uses
- Plugins - reflection - axpress reacting to itself
- Queries - 'one liners'
- Apps - music player, video player, etc
- Compiler - the details
- Evaluator
- SPARQL - the interface to SPARQL endpoints
- Data Structures - how might other storage methods and data structures work (for improved efficiency)
- GUI - Whats it look like?
- Language Interface
- Ideas
- Storage
- Problem Queries
- Learning
- Growing
Optimizing for specific usage patterns
In the case of the wiki example, to store each revision of the page with a date, means that to recall the most recent version of a page in SPARQL requires the endpoint to find all versions of that page, and look for the most recent one. Depending on the implementation, it may even sort the entire version history by date, then pull off the top. This raises the issue of if it is possible to detect and match on the way information is commonly used. In this case, only the most recent page is by far the most commonly retrieved, so why not keep a 'pointer' to the most recent page? The problem is essentially how to translate these queries and sets of queries into code which acts in ways similar to the ways that a programmer who was implementing them by hand would program them. Can these patterns be detected? And is it easy enough to do that it is worth doing?
In cases where speed is extremely important (like real-time audio and video), for now, the pieces which need to be fast should not be written with this language.