This updates the ERIS-FS format by reducing it to an index that carries references to file contents which are themselves encoded using ERIS.
Imho this simplifies the format considerably and allows much simpler parallelization. The requirement for a random-access decoder has been lifted. Encoding/decoding of individual files can be parallelized. Implementation in general is simpler as there is no need to count block offsets or do any padding.
Support for compression has been dropped. It can be added later with additional entry types. Removing compression has the advantage that this format can be implemented much easier on constrained platforms (e.g. browsers). It also makes creation of ERIS-FS file system much easier in GUI applications.
This change was motivated by considerable performance issues when working with real-world use-cases (Guix substitutes).
This is a breaking change, the version number in the index has been incremented.
This updates the ERIS-FS format by reducing it to an index that carries references to file contents which are themselves encoded using ERIS.
Imho this simplifies the format considerably and allows much simpler parallelization. The requirement for a random-access decoder has been lifted. Encoding/decoding of individual files can be parallelized. Implementation in general is simpler as there is no need to count block offsets or do any padding.
Support for compression has been dropped. It can be added later with additional entry types. Removing compression has the advantage that this format can be implemented much easier on constrained platforms (e.g. browsers). It also makes creation of ERIS-FS file system much easier in GUI applications.
This change was motivated by considerable performance issues when working with real-world use-cases (Guix substitutes).
This is a breaking change, the version number in the index has been incremented.