Pangram tests do not match exercise specification

[SiriusStarr]

Issue #824 and PR #827 added tests to the pangram exercise to check handling of non-ASCII characters. However, the specification for the exercise in exercises/pangram/README.md states:

Input will not contain non-ASCII symbols.

Presumably this should be removed from the exercise specification if solutions that cannot handle non-ASCII input are going to fail.

[sshine]

You are very right. Thanks for reporting this.

Unfortunately, deviating from the canonical description in README.md is not something that we currently do. Doing so has some implications that have not been described in full anywhere; the biggest drawback is that syncing with the canonical description becomes gradually more difficult if we diverge from it.

In another recent example, resistor-color-duo, I considered to deviate from the canonical description (#865). Here I submitted exercism/problem-specifications#1605 instead in the attempt to fix this upstream rather than track-locally. The difference between that case and pangram is that resistor-color-duo is a new exercise, and the problematic sentences were added after the first solution was ported to Haskell. Unlike pangram which is a quite old exercise; fixing this upstream would be possible if we argue that "restricting oneself to ASCII should be treated as an optional thing, and the problem description should be compatible with tracks that choose to support non-ASCII input (but still regard a pangram as a-z)."

Unfortunately, problem-specifications PRs are currently on hold until a structural problem is resolved.

So I suggest we live with this discrepancy for now, leave this issue open as a reminder, and address it again when something is born from the ashes of the problem-specifications repo.

[sshine]

Since problem-specifications has been re-opened, requesting to remove this line from the canonical exercise description is now an option.

I don't know if people agree on this (agreement was the reason the repository was closed for a year), but an alternative for the many languages practiced on Exercism where non-ASCII input would be a severe complication, one could encourage to leave it as a track-specific ASCII-only hint. (The point is that we can't do the opposite without contradicting ourselves.)

@SiriusStarr: Are you interested in attempting this?