Separate JSON and hash file operations are inevitably not atomic #515

Activity

1. arkjedrz

   added the

   Task

   issue type on Oct 16, 2026
2. github-project-automation

   added this to PER - Persistency FTon Oct 16, 2026
3. github-project-automation

   moved this to Backlog in PER - Persistency FTon Oct 16, 2026
4. arkjedrz

   self-assigned this

   on Oct 20, 2026
5. arkjedrz commented on Oct 21, 2026

   

   arkjedrz

   on Oct 21, 2026

   Last edited by arkjedrz

   ContributorAuthor

   More actions

   File format proposal

   I propose the following, sizes are in bytes.

   [ header : 4 | version : 4 | checksum : 4 | data : n ]

   • header: file header containing a magic number, I propose SKVS == 0x534B5653. File format detection only.
     • Should be a constant, always.
   • version: representation version.
     • Should be changed on data representation or file format change.
     • It is not a version of consumers data!
   • checksum: checksum of data that follows. Size might be increased in the future, then a bump to a version should occur.
   • data: data stored continuously.

   header and version layout should always be the same.
   The rest can be modified with a version bump and either ensured compatibility or reported version mismatch.
   E.g., version 2 might introduce CRC64, which makes checksum part 8 bytes long. Version 3 might introduce a new flags section placed before checksum.

   I'm not certain if additional data length field should be added to avoid unrestricted read.
6. arkjedrz commented on Oct 22, 2026

   

   arkjedrz

   on Oct 22, 2026

   ContributorAuthor

   More actions

   Discussed and agreed 22.10.2026