# RMO83 — The current input-error check

**Result in one sentence:** The perpendicular model uses the refined physical
check automatically; the observer does not need to choose between checking methods.

## How to use it

Open the full QuickLook and click **Try the perpendicular model with errors**.
Choose an input-bound factor and click **Load example and view result**. The
short sentence, speed plot, physical checks and downloadable result PDF refer
to the current method. All three demonstrated nonzero error settings support
the fast-shock type for this constructed model under exactly perpendicular
geometry and the stated single-front ideal-MHD assumptions.

With the Python connection active, editing the parameters and clicking
**Calculate these inputs with Python** uses the same assessor. Opening the
HTML as a downloaded file provides the current checked examples and downloads;
it does not start a Python service. The interface states which mode is active.

The relation imposed by conservation concerns the physical states. Errors of
separate measurements may nevertheless be independent. Keep the measured values
and their uncertainty; the method does not silently replace them by compatible
values. The perpendicular refinement still requires a checked nominal anchor,
exactly zero normal magnetic field and zero width for that component. It does
not fit an inconsistent centre or infer an unknown field angle.

## What is checked

The method first checks the exact central state and the applicable type
inequalities. For the perpendicular model, the entropy bound also uses the
relation between the states imposed by conservation. This is a sufficient
conditional check, not a complete search of every possible MHD solution.

The result applies to conservation-compatible states inside the supplied
ranges. It does not assert that every combination of independently varied
inputs is a physical solution. Nor does it establish that the constructed
model describes an observed solar front. Unknown geometry and an inconsistent
central state still require separate work.

The input-bound factor scales the saved widths: density and pressure use a
fraction of their central magnitudes; velocity, front speed and tangential
magnetic components use the factor times max(1, absolute central value) in the
stated normalized units. Both normal magnetic components and their widths are
zero for this model. The factor is not a uniform percentage or a probability.
The actual input ranges are available in the parameter section and JSON.

There is one current method in QuickLook. Its result, speed plot, physical
checks and per-input PDF refer to the same assessment. The separate exact
central-state checks remain available. Development comparisons and original
records are retained in the complete project archive; they are no longer an
additional reading path in the current method Help.

## Validation retained from RMO82

- 25 protected loopback HTTP/worker checks: saved/fresh assessments and PDFs
  agree for all four P02 settings; exact input identity, revision and hashes
  are preserved. A changed width changes the evaluated bound. Unknown angle,
  inconsistent centre and additional failed inequalities are not promoted.
- 15 controlled-DOM checks exercise all 24 active examples, exports, edits,
  import feedback, stale-result rejection, response identity and service errors.
- Per-input PDFs retain the actual assessment method, conditional result,
  geometry, independent audit, plotted ranges and input checksum. The new
  representative PDF was rendered and visually reviewed.
- Scientific modules, earlier numerical output files, figures and the RMO76
  solar audit are unchanged. RMO82 integration preservation is recorded in RMO82_bundle_manifest.json.

RMO83 changes only the method explanation and its presentation. Its preservation
check confirms that scripts, saved inputs/results and current plots/PDFs are
unchanged, and that the full RMO82 page can be recovered exactly. No new
scientific calculation is claimed for this presentation change.

These controls do not claim new native Chrome, Windows Python or public-hosting
acceptance. No public site has been deployed. Numerical results remain model
demonstrations, not independent classifications of observed solar EUV fronts.

## Next scientific dependency

Revisit the existing solar-event audit with matched geometry, plasma-frame
velocities, thermomagnetic constraints and error provenance. The model result
does not supply missing solar measurements. Uncertain angle and noisy-centre
feasibility remain distinct limitations.