Problem

The PROCESS_ISOLATION.md document currently contains an analysis claiming that JGDMS TlsRMIServerSocketFactory has two "correctness gaps":

1. Missing equals() / hashCode() — described as required by the RMI specification
2. Missing Serializable — described as needed for RMI Registry stub distribution

However, these claims require clarification:

Regarding equals() and hashCode():

• The RMI specification uses "should" (recommendation), not "must" (requirement)
• For a stateless factory with no instance state, Object's default equals() and hashCode() (based on reference identity) are perfectly adequate and appropriate
• Without mutable or comparison-relevant state, custom implementations are unnecessary

Regarding Serializable:

• There is no specification requirement that RMIServerSocketFactory implementations be serializable
• The factory is primarily used by the RMI Registry locally, not transmitted across wire boundaries
• No specification language mandates serialization for factories used in a Registry context

Solution

Update the "Analysis: JGDMS TlsRMIServerSocketFactory vs DirtyChai SslRMIServerSocketFactory" section in PROCESS_ISOLATION.md (around line 2655–2707) to:

1. Revise the equals() / hashCode() claim: Remove or soften the language stating these are "missing" requirements. Clarify that for a stateless factory, Object's default implementation is the correct choice. Note that if the factory gains configurable state in the future (like SSL context or cipher suite preferences), then equals() / hashCode() would become necessary at that point.

2. Revise the Serializable claim: Remove the language stating this is a missing requirement. Clarify that serialization is not mandated by any specification for RMI Registry socket factories, which are used locally, not transmitted.

3. Update the summary table: Adjust the entries for equals() / hashCode() and Serializable to reflect that these are not gaps but rather appropriate design choices for a stateless factory.

4. Preserve the architectural verdict: The comparison showing that JGDMS TlsRMIServerSocketFactory is superior due to Subject integration and principal-keyed policy remains correct and should be retained.

The following is the prior conversation context from the user's chat exploration (may be truncated):

User: Regarding TlsRMIServerSocketFactory, I couldn't find any reference stating it must implement equals and hashcode, only that it should, there was no mention of implementing Serializable. It is designed primarily for the RMI Registry, it doesn't need to be Serializable. Additionally, it is stateless and has no state for equals comparison, hence, Object equals and hashCode are sufficient.

Assistant: I'll help you investigate the TlsRMIServerSocketFactory and its relationship to equals, hashcode, and Serializable. Let me start by examining the repository and the specific file you referenced.

Assistant: I've reviewed the PROCESS_ISOLATION.md document and the code context. Your clarification is well-founded. Let me explain what the analysis currently states and address your points:

Current State in PROCESS_ISOLATION.md

The document contains an analysis section titled "Analysis: JGDMS TlsRMIServerSocketFactory vs DirtyChai SslRMIServerSocketFactory" which claims that TlsRMIServerSocketFactory is missing two things:

1. Missing equals() / hashCode()
2. Missing Serializable

Your Clarification

You're correct on all three points:

1. On equals() and hashCode():

The RMI specification uses "should" (a recommendation), not "must" (a requirement). Your point is valid: without custom state to compare, using Object's default equals() and hashCode() is perfectly sufficient. Both rely on reference identity, which is appropriate for a stateless factory.

2. On Serializable:

You're right — there is no specification requirement that RMIServerSocketFactory implementations be Serializable. The factory is primarily used by the RMI Registry locally, not transmitted across the wire. The DirtyChai code shows that SslRMIServerSocketFactory does implement both equals() and hashCode(), but neither it nor TlsRMIServerSocketFactory strictly requires Serializable for RMI Registry use.

3. On statelessness:

Since TlsRMIServerSocketFactory is stateless (no instance fields), Object's default equals() and hashCode() implementations are not just sufficient—they're the correct choice. Custom implementations would only be needed if there were mutable or comparison-relevant state.

Suggested Action

The analysis in PROCESS_ISOLATION.md should be updated to reflect this clarification:

• Remove or revise the claim that equals() / hashCode() are "gaps" for a stateless factory
• Remove the claim that Serializable is required for RMI Registry use
• Optionally note that if `Tls...