How we will testing V2 (2024 Feb-May)

[heavily WIP]

The testing of the V2-Pokfield happens over 2024 Feb-May, in three phases:

  1. Character defaults (Feb 1 – Mar 1)
  2. Full texts (Mar 1 – Apr 1)
  3. Culture / Ecology (Apr 1 – May 1)

1. Character Defaults

Each character had been assigned a default reading, referenced from a data file table1_vf_2024-01-15.json. These had been checked in the lead up to v2.0.20-alpha but more should be done.

1.1 Frequency list

The plan here is to prepare a font-rendered document with the top n most frequently used characters. n depends on who is available; I think I can do 6,000. Jon will generate and provide a spreadsheet which can be used to annotate.

What we need to do with each character is to examine whether it is the “right” default — I put right in quotes here because this is a matter of choosing the least evil.

In general, the project seeks to be pragmatic. We anticipate users mostly interested in preparing vernacular-colloquial teaching material, and seeks to

1.2 Verb list

1.3 Simplified-preference list

2024 Feb

Character defaults

In February, I have checked once again the top 8,000 frequently used characters for their default readings, and corrected them. Each correction was accompanied with a word/compound patch list, and these were rolled into 2.5.

Verb / Simplified preference list

There was only one verb that required a “list usage treatment” 彈; and the number of Traditional characters whose Simplified reading is preferred turns out to be very few.

Full text benchmark

Parallel with the above, 18,000 characters of text, from a variety of types, had been bench-marked. Modern texts are 99.7% accurate with 2.4 and below, and 99.8% with 2.5 and above. Classical texts, like Tang poems, which are precisely what the font is tuned against, nonetheless still score ~98%.

Compatibilities

A variety of issues with Windows (requiring Microsoft specific OS/2 language codePages; family name length) had been identified.

On Mac it had seen uses in iWorks, MuseScore, Chrome, (B/W) Premiere burnt captions. Usage as video subtitles is very exciting.

I’ve worked with VectorStyler, a design application that works cross-platform Win/Mac, to get font rendering bugs squashed; their 2024-03-10 update will fully support Canto Font’s demanding requirements.

Some .otf builds have been demonstrated to work on Kindles.


My overall progress had exceeded my expectations, and I anticipate 2.5 to be stable to release. There may be minor features :sunglasses: to be corrected or added, but the font is fully functional and stable.

What remains untested / unknown are how the font behaves:

  • in different browsers,
  • on Android (how to bring in custom font?)
  • on other eReaders (Kobo, and…?)
  • on Linux