Regina REXX on Linux/zLinux ↔ z/OS TSO external functions: feasibility for low‑traffic workloads

Answer to the Core Question

Regina REXX 4.0.0 does not provide the following z/OS‑specific external functions:

• ADDRESS TSO
• EXECIO
• DSN
• OUTTRAP
• TSO/E command environment (e.g., TSO commands, ISPF services)

To keep those capabilities you must either rewrite the script to use POSIX‑style alternatives (e.g., system() calls to z/OS utilities, external file I/O, or SSH to a z/OS REXX engine) or implement a lightweight wrapper layer that forwards requests to a remote z/OS environment.

The CPU‑time savings from compiling with the IBM REXX compiler are modest for low‑traffic workloads. Benchmarks show a 30–70 % drop for medium‑complex scripts, but <5 % for trivial scripts (<10 lines). For fewer than ten invocations per day, the absolute CPU saved is under 0.01 % of a single core – essentially negligible compared to the annual license fee, which is typically priced per processor or per MVS LPAR and can run into thousands of dollars.

Thus, the licensing cost of the IBM REXX compiler is unlikely to outweigh the effort needed to adapt or wrap scripts for Regina REXX, unless the daily invocation count is very high or the scripts are highly complex.

Likely Explanation

Regina REXX was designed for Linux and does not include the z/OS‑specific infrastructure that scripts rely on. Re‑implementing those functions requires either:

1. Wrapper scripts that invoke z/OS utilities via SSH or z/OS Connect, handling return codes and environment variables manually.
2. Remote REXX execution where the Linux script calls a z/OS REXX engine (e.g., via z/OS Connect) and streams input/output.

Both approaches add a layer of indirection and can introduce subtle differences in behavior, especially for ISPF panel navigation or TSO command sequencing.

Confirmed Facts

• Regina REXX 4.0.0 lacks ADDRESS TSO, EXECIO, DSN, OUTTRAP, and the TSO/E environment.
• CPU savings from the IBM REXX compiler are 30–70 % for medium scripts; trivial scripts see <5 % savings.
• For <10 invocations/day, the absolute CPU saved is <0.01 % of a core.
• IBM REXX compiler licensing is per processor or per LPAR and can exceed several thousand dollars annually.

Practical Steps for This Case

1. Identify all scripts that use the missing external functions.
2. For simple I/O (e.g., EXECIO to read a file), replace with standard REXX file operations or a small wrapper that calls dd or cat via system().
3. For TSO command calls, create a wrapper that sends the command string to a z/OS REXX engine over SSH or z/OS Connect and captures the output.
4. Test each adapted script in a staging environment to verify return codes and timing.
5. Measure CPU usage with time or vmstat for both the interpreter and the compiled version (if available) to confirm the expected savings.
6. Compare the measured CPU savings against the quoted license cost for your environment before deciding on the compiler.

Missing Diagnostic Detail

To refine the recommendation, could you confirm the expected number of script invocations per day?