Sergey Bykov, principal architect of Microsoft Orleans and now at Temporal, has a confession: after over a decade pioneering virtual actor implementations, he deliberately avoids the term "actors." This revelation, sparked by David Fowler's viral tweet, exposes a fundamental identity crisis in distributed systems design. Bykov argues that the term's baggage creates unnecessary friction—and suggests it's time for a rebrand.

The Conflation Minefield

When Orleans launched in 2015, purists dismissed it as "not real actors" despite its peer-reviewed foundations. Bykov identifies four core conflations that fuel such debates:

  1. Distributed vs. Local Actors:

    "Many developers use actors as a concurrency mechanism... but distributed actors live in a hostile environment of network failures and latency."
    Single-process actors (common in Java) solve different problems than cloud-scale systems like Orleans.

  2. Supervision Trees ≠ Actor Model:
    Erlang/Akka's supervision hierarchies became mistakenly synonymous with actors. Orleans' automatic lifecycle management—optimized for cloud resilience—required constant justification against this expectation.

  3. Messages vs. RPC:
    Traditional actors favor one-way messaging, viewing request-response as dangerous. Orleans embraced async RPC with timeouts and reentrancy, leveraging .NET's async/await for safe concurrent calls.

  4. The 'Become' Fallacy:
    Some insist actors must explicitly change behavior via methods like Become. Bykov counters that internal state transitions (e.g., IoT device twins) satisfy the model's requirements.

The Naming Barrier

"Actor" fails as an intuitive metaphor, Bykov argues. Most developers envision Hollywood—not encapsulated stateful entities. When presenting Orleans, he now describes grains as:

"Objects with stable identities living in a cluster, accessible via async method calls."

This object-oriented framing resonates better than battling preconceptions. Cloudflare's "Durable Objects" and Reuben Bond's "Cloud Native Objects" (CNOB) are more effective labels, he suggests.

Why Terminology Matters

Naming isn't pedantry—it affects adoption. The Actor Model's principles (encapsulated state, async messaging) power mission-critical systems, but linguistic baggage obscures its value. As Bykov notes:

"Two hard things in computer science: naming things and cache invalidation. I believe [naming] is part of why actor frameworks aren't more popular."

For cloud-native builders, the lesson is clear: focus on capabilities (stable identities, failure isolation, horizontal scaling) rather than ideological purity. The future belongs to concepts that describe what they do—not academic legacy.

Source: Based on "The Curse of the A-Word" by Sergey Bykov (Temporal).