Endevor versus Git / GitHub

This post details the technical problems, hidden expenses, and the legality

of moving an Endevor source code management system to Git / GitHub.

The hypothetical zOS production environment is comprised of 4,000 z/OS application

COBOL, PL/I, and HLASM Endevor source code modules that require yearly updates.

Distributed version control works well for open-systems and cloud-native

software. However, putting Git in place for z/OS COBOL, PL/I, HLASM applications

creates major data errors, compile issues, and oversight gaps.

These structural flaws must be addressed before even contemplating test projects.

1. Character Encoding & Data Losses (EBCDIC vs. ASCII)

Git runs on ASCII or UTF-8 text formats, while z/OS uses EBCDIC. Putting Git

between these systems creates text translation flaws:

* Hexadecimal and Packed-Decimal Data: zOS COBOL, PL/I and Assembler code

  hold embedded hex constants and packed-decimal code. Standard text tools

  cannot tell the difference between plain text and non-display code constants.

* Conversion Errors: Changing source files from EBCDIC to ASCII for local

  editing, and then back to EBCDIC for compiling, breaks character mappings.

  One broken character will pass compiler checks but trigger S0C7 data crashes

  or silent data errors during batch processing runs.

2. Operational Risks of Shared Code Merging

Distributed SCM tools use side-by-side branches and automated text merging

instead of locking code files. This setup removes proven change-management

safety steps:

* COBOL Auto-Merge Flaws: Git merges files line-by-line as raw text. It does

  not understand COBOL paragraph layouts, copybook setups, or file links.

  Double edits to a large COBOL file can pass auto-merges while breaking

  column spots, variable rules, and program logic.

* Loss of Overwrite Protection: Endevor’s "Check-Out and Lock" rule stops

  developers from overwriting linked code files. Moving to automated branching

  creates code clash risks across the 4,000 modules. Scheduling changes

  cannot fix this, leading to more system downtime.

3. Extra Work & Server Costs

Moving to a local Git setup does not cut down admin work; it just moves the

work to higher-paid IT jobs:

* Staffing Needs: Keeping offline, fire walled code servers like GitHub

  Enterprise Server or GitLab Self-Managed running safely takes 1 to 2

dedicated systems admins. They must handle server crashes, daily backups,

  security updates, and login syncs.

* Estimated Costs: This additional labor cost of $200,000 to $350,000 annually,

in addition to software license expenses.

4. Technical Skill Gaps

Standard new age workers do not have the specialized system skills to run

mainframe compile pipelines:

* Systems Programming Limits: Mainframe systems teams focus only on OS updates,

  software fixes, and system uptime. They do not help build or fix application

  compile pipelines.

* Contractor Knowledge Gaps: Standard new age contractors do not know how

  Partitioned Data Sets, JCL compiles, or mainframe link-edit rules work.

  Workers who know both Git steps and Endevor setups are very hard to

  find. Hiring them requires costly help at $250+ per hour or long internal

  training costs -- with no guarantee that a trained asset will not seek other jobs or assignments