cs.vt 20190221

A content hash based data store with a filesystem layer, using variable sized blocks, arbitrarily sized data and utilising some domain knowledge to aid efficient block boundary selection.

A content hash based data store with a filesystem layer, using variable sized blocks, arbitrarily sized data and utilising some domain knowledge to aid efficient block boundary selection.

NOTE: pre-Alpha; alpha release following soon once the packaging is complete.

The package provides the vt command to access these facilities from the command line.

This system has two main components:

• Stores: storage areas of variable sized data blocks indexed by the cryptographic hashcode of their content
• Dirents: references to filesystem entities containing hashcode based references to the content

These are logically disconnected. Dirents are not associated with particular Stores; it is it sufficient to have access to any Store containing the required blocks.

The other common entity is the Archive, which is just a text file containing a timestamped log of revisions of a Dirent. These can be mounted as a FUSE filesystem, and the vt pack command simply stores a directory tree into the current Store, and records the stored reference in an Archive file.

See also the Plan 9 Venti system: (http://library.pantek.com/general/plan9.documents/venti/venti.html, http://en.wikipedia.org/wiki/Venti) which is also based on variable sized blocks.

Function totext(data)

Represent a byte sequence as a hex/text string.