Incremental dumps/TODO

Required for production deployment

• code review
• integrate with xmldumps-backup
• reasonable handling for file-related errors (file not found, access denied, etc.); right now, the application just crashes
• Tracking bug

Later

• better compression for metadata and indexes
  • compress indexes using varints (or LZMA?)
  • add index for named users
  • compress list of revision ids of a page using varints
  • store parent id as a varint difference from revision id
  • add index for namespaces
• see if it's possible to tweak LZMA (language-specific Huffman tables?)
• library for binding with other languages