Base64 Lamport timestamps
The package implements Base64 string Lamport timestamps. First described in "Time, clocks, and the ordering of events in a distributed system" by Leslie Lamport these timestamps are designed to track events in a distributed system. The paper's primary inspiration was the special theory of relativity. It describes a model of time based on sequential communicating processes each having its local clocks only, no "newtonian" global universal clocks.
Hence, every timestamp has two components:
- monotonically increasing "clock" and
- process identifier.
These days, Lamport timestamps are used everywhere, starting from multicore CPUs all the way to world-scale distributed systems.
This implementation deals with base64 string based timestamps of variable length. Those are handier and more flexible than C-style fixed-width binary formats. Base64 can be used inside URLs (path/fragment parts), logs, arbitrary databases, etc. Importantly, Base64 timestamps are human readable.
The clock component may be a simple sequential counteror a timestamp from a physical clock (data/time) or something inbetween or combined. (There are lots of fine details here.) Classes:
- TestClock (sequential counter)
- MinuteClock (timestamp + sequence number, exact to a minute)
- SecondClock (timestamp + sequence number, exact to a second)
- AdabtableClock (timestamp + sequence number, long enough to guarantee uniqueness and monotonic growth)
Base64 timestamps obey alphanumeric order, that's why the package employs its own variety of base64. In common base64 variants, numeric order does not match alphanumeric order (i.e. the order for numbers and their base64 serializations differs).
swarm base64: 0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ_abcdefghijklmnopqrstuvwxyz~
The package also implements two varieties of version vector based on Base64 Lamport timestamps:
- regular version vector (VVector, which is actually a map) and
- arrival order preserving version vector (OrdVVector).
See tests for usage examples.