Albert Zak

Download the original poster as PDF (9MB) >

Presented at the Symposium on Reliable Distributed Systems (SRDS) 2022 PhD track.

Interactive Programming
as an
"Operating System"
for Resilience in Distributed Systems

Goal#

find data model + editing ui

• inspect – see the state
• experiment – modify the state transiently
• evolve – experiment with real data, but safely manage side effects

User interactions (front end)#

• plain text/s-expressions
  semi-structural editing
• custom specific ad-hoc "lenses"
  inline editor visualizations
• feature flags and tracing
  probes via sigils

Data model (back end)

immutable log of atomic facts

• history/time-aware
• represent both code (top level defs)
• and state (ephemeral and persistent)
• + reifies changes to both with provenance

EAVT (entity, attribute, value, time)

• access path independent 6NF
• E, A, V fully indexed, time for auditing purposes only
• save now, query later

capabilities and event handlers

• confine side effects for safe experimentation

Why?#

• situated systems (irregularity, information, continuous use, humans, change over time)
• safety + security w.r.t ambient authority

Related work#

Lisp machines, Smalltalk, Multics, Erlang/OTP, Emacs, REPLs, Notebooks, Datalog, Clojure, Datomic, XTDB, APX, STEPS, Eve, Dark, Unison, Mu, WASI

plain data, immutable

plain text, embellished

effects and capabilities

library, not a framework