The website still bills NilPy as experimental; the docs already say otherwise

RESOLVED 2026-08-27 — shipped as 4beadde, verified from the public internet

Verified by frank2-af with a plain curl, independently of the origin-side check:

<meta name="description" content="PXX is a from-scratch Pascal, C and Python
  compiler — self-hosting, cross-compiling to several CPU targets, highly FPC-
  and C99-compatible, with nilpy compiling Python to native binaries.">

/ and /about/ carry no remaining staged / in development / never. The one out of scope still on /compliance/ is the C row — "C99 — GCC extensions out of scope" — which is correct and unrelated.

ianweb made the nilpy fidelity claim harder, not softer: "Python syntax only" became CPython observable-behaviour parity with divergences filed as bugs, lifted from docs/targets/nil-python.md. That is the right call and worth noting — the page had understated the ambition and overstated the limits at the same time, and it gives the deep NilPy queue its honest reading: an exacting reference target, not a fragile frontend.

THE GREP LESSON — it was SIX locations, not four

The table below listed four. There were six. The miss was about.md:59, a footnote under the About page's footnote list, carrying both "Staged —" and a second independent copy of the library-ecosystem claim.

Why it evaded the sweep: the four were found by grepping the wordings already known — experimental|staged|in development. The sixth phrased the same claim differently, in prose reachable only through a footnote link. ianweb found it by grepping the claim-space across all of pxxweb/content/ before editing — staged|experimental|in development|out of scope|never — which cost seconds.

The rule: grep for the CLAIM, not for the WORDING you already found. A string sweep answers "where does this sentence appear", which is a different question from "where does the site say this thing" — the same correct-reading-of-the-wrong-instrument shape that cost this lane an afternoon, in search rather than in measurement. Prose restates; only the claim is invariant. Corollary: sweep before editing, not after, because the cheap pass is worthless once you have already fixed the instances you knew about.

Owner, 2026-08-27:

"on website, subject compliance, we still list nilpy as experimental. i think that's no longer fair - python is first class citizen by now, despite the list of open bug tickets."

Fix lands in the private ~/pxx-website repo, not this checkout.

The four places, measured

config.py matters more than its size suggests: that string is the <meta name="description">, the og:description and the twitter:description on every page, so it is what search engines and link unfurlers quote. Fixing the compliance table alone leaves the stale claim on the most-syndicated surface the site has.

The docs already disagree, which is the strongest evidence here

docs/targets/nil-python.md:10, in the pxx repo, published by this same site:

It is a mainline frontend, a peer of Pascal and C rather than a research path: it has its own test gate (test-nilpy), which must be green — along with a byte-identical self-host and the cross-target builds — before any change lands. BASIC, Rust and Zig are the experimental frontends; Nil Python began there and no longer is.

CLAUDE.md says the same: N is mainline (peer of C, not experimental like R/Z), with its own carved-out files and a gated suite. So the site currently contradicts both its own docs and the project's charter, on a page whose whole subject is telling people what is true.

Why "open bug tickets" is not a counter-argument

The owner pre-empted this and he is right. Every mainline frontend carries open tickets; Pascal and C carry more, and neither is billed as experimental. "Experimental" is a statement about commitment and gating, not about defect count — R and Z are experimental because they are unranked, optional and ungated, which is precisely what NilPy is not.

Scope

Change the four strings. Keep the class of claim honest — "mainline", "gated", "peer of Pascal and C" are all supportable from the gate; do not upgrade the language further than the docs already go.

A SECOND claim on the same line, NOT covered by the owner's call

compliance/overview.md:14 also says:

Python syntax only — the CPython library ecosystem is out of scope (pxx is a compiler, not a Python runtime)

That may be stale too — feature-nilpy-stdlib-coverage-gaps-measured, feature-nilpy-thirdparty-libraries-as-targets and the SQLite/tkinter work all point the other way, and CLAUDE.md's rule is that NilPy is upward compatible with CPython ("if code works on CPython, it must work on NilPy"), which is not the posture of a project treating the library ecosystem as out of scope.

Do not change it on this ticket. The owner spoke to "experimental" only. Flagged so whoever edits line 14 notices they are standing next to a second claim, and asks rather than tidies.

SECOND ITEM, same page, same edit pass — soften "never" on C++

Owner, 2026-08-27:

"oh, and same page lists c++ as 'never'. i'd say, that's still accurate. just, never say never."

The substance is endorsed and does not change. C++ is not planned, the reasoning below the table is sound, and nothing here promises a C++ frontend. Only the absolute goes.

Two places, and they must move together — softening the table row while the prose underneath still says "deliberately never" reads worse than leaving both:

Suggested shape, not prescribed wording: Not planned in the status cell, and a heading and paragraph that give the same reasons in the present tense — the grammar is huge and context-sensitive, the build model drags in a heavyweight toolchain, the cost does not justify the payoff today. That says everything the current text says about intent while not committing the project to a position it would have to eat later.

Why this is worth doing rather than pedantry: the page's own opening sentence promises "the honest map". An absolute is a claim about the future, which is the one thing a status page cannot measure — and it sits three rows below a Never on a language that has an experimental frontend two rows up. The credibility cost of an unnecessary absolute is paid the day it changes, and this project has already had to correct one over-strong claim class (see CLAUDE.md's byte-identical discipline). Same failure, smaller stakes.

Note the status cell also uses class="cstat exp" — the experimental colour — for Never, which is a class mismatch rather than a style choice. Fix or leave, but notice it.

Lane note

The owner called this "a small track D notice". Filed as W because D owns docs/** in the pxx repo and these files are pxxweb/content/** in the website repo — the letter follows file ownership, which is what the lanes are for. The work itself is D-shaped prose. Noted rather than silently reclassified.

Log

• 2026-08-27 — resolved, commit f6f9879d9.