Render every current JSP and Jakarta tag/EL binding through isolated JSP engines #120

Reviewed 2026-09-25 (America/Los_Angeles) against main at bd249f5. Execution order and cross-issue ownership: #169. Batch 03.

This scope replaces the dated implementation prescriptions in the original report and earlier comments; linked historical evidence remains useful but must be rechecked before implementation.

Remaining coverage gap

#167 now validates TLD declarations, reflection, invocation and source parity; it does not execute a JSP compiler. #151 added JSON and #168 proposes XML 1.1, so the old fixed 60-binding count is obsolete. Keep this as parser/container integration coverage rather than duplicating reflection tests.

• Use isolated test-only Jasper/container fixtures compatible with each adapter's published API baseline. Select maintained versions and review advisories at implementation time; do not put a conflicting newer JSP API on existing unit-test classpaths.
• Generate/verify coverage from the actual packaged basic and advanced TLDs, including JSON and XML 1.1 once merged, and fail if any supported binding is untested.
• Render hostile text, null/missing values and EL coercions; compare exact emitted bytes to the intended facade contract and test rejection of tag bodies/missing required attributes. Avoid treating the facade oracle alone as evidence that a browser context is safe.
• Cover both Jakarta and legacy javax adapters with the correct engines, or record a specific remaining legacy gap before closing; keep dependencies test-only and libraries at their existing Java/provided API baselines.
• Run the Docker-free checks under normal verify without changing shipped artifacts. Preserve Add isolated packaged-consumer compatibility CI across Java 8–25 #162 and Guard facade mappings and JSP/Jakarta parity #167's existing checks.

Sequence after #131/#168 to avoid rebasing the surface twice. #93 owns whether/how the separate browser app is retained; it must not be retired merely because these byte-comparison tests pass.