Libertinus: Glyph list standards

[iandoug]

Hi

Somewhere between a rant and a mind dump. Inputs and counter-arguments welcome.

See attached analysis of various "wide glyph support" fonts. The analysis is a work in progress, there may still be bugs (eg around the private use areas, which I'm trying to ignore).

Fonts are:
Arial Unicode
Bablestone Roman
Brill
Code2000
Dejavu Math
Junicode
Libertinus Math
PalemonasMUFI
Quivera
STIX
Symbola
TITUS Cyberbit

Support for various "glyph list standards" is spotty.

After some thinking, I've come to the conclusion that these standards have their own agendas, which are not entirely relevant to fonts with broader academic use ideals.

Eg.

1. Google wants glyphs typically used on web.
2. Adobe and Linotype want to sell fonts to as many different countries as possible, but at the same time will want to sell the Latin version and the Pi version, and not put all the glyphs in one font.
3. Microsoft is focused on glyphs needed in Windows, not necessarily documents written on Windows.
4. SIL is focused on characters needed for their literature, aimed at under-serviced languages, with some historic middle-Eastern additions.
5. MUFI is aimed at dealing with Medieval documents.

I added the STIX glyph list as a "standard" to see what popped up.

Comparing glyph counts, per standard:

GCore : 215 Googie Core
AL1 : 229 Adobe Latin 1
AL2 : 250 Adobe Latin 2
SIL : 295 SIL Recommended Glyph List for new fonts
AL3 : 331 Adobe Latin 3
GPlus : 561 Google Plus
AGLFN : 586 Adobe Glyph List for New Fonts
W1G : 587 Adobe World Glyph List 1
AL4 : 617 Adobe Latin 4
WGL4 : 657 Windows Glyph list 4
GPro : 677 Google Pro
GExpert : 699 Google Expert
MUFI 3.0 : 756 The Medieval Unicode Font Initiative 3.0
MUFI 4.0 : 782 The Medieval Unicode Font Initiative 4.0
MUFI 4.n : 785 The Medieval Unicode Font Initiative 4.n
W2G : 849 Linotype World Glyph List 2
AL5 : 1307 Adobe Latin 5
STIX : 3300 STIX Mathematics

Fonts fail to meet standards at different points. For example, Arial Uni initially fails WGL4 because it lacks the Cyrillic version of È.

So ... I think these "standards" need a different approach, because one size does not fit all. I see a practical system along these lines:

1. Humanities
2. Sciences

The humanities could be further subdivided, by continent and/or time period. So a font may support American languages, but not African.

Not sure if trying to divide up the sciences makes sense, there is a lot of overlap and re-use.The STIX list is a good starting point, but may also be too high a standard, given the glyph count.

Comments welcome.

Cheers, Ian

[Crissov]

• Does Libertinus Serif fare any better than Libertinus Math?
• Which sets would be low hanging fruit to conform to for Libertinus?

[iandoug]

@Crissov

1. No. I thought Math was a superset of serif?
2. That's up to leadership :-)... while I work on a better "report" to answer such things, here's a table export comparing Math, Serif and Sans. I guess you mean "which needs fewest glyphs", that's the bit I need to work on.

Will revert.

Cheers ,Ian
fontcompare.zip

[iandoug]

Okay ... so actually there is some low hanging fruit ... two glyphs gets you to Adobe Latin 1 and 2, another 2 to Adobe Latin 3,
then Commercial Minus and a few currencies takes care of Google Plus, Pro and Expert. Adobe Latin 4 also almost done at that point.

Lib Math extract. See attachments.

Adobe Latin 1

U+FB01 Latin Small Ligature Fi
U+FB02 Latin Small Ligature Fl

Adobe Latin 2

U+FB01 Latin Small Ligature Fi
U+FB02 Latin Small Ligature Fl

Adobe Latin 3

U+20BA Turkish Lira Sign
U+20BD Ruble Sign
U+FB01 Latin Small Ligature Fi
U+FB02 Latin Small Ligature Fl

Adobe Latin 4

U+20A1 Colon Sign
U+20A6 Naira Sign
U+20B2 Guarani Sign
U+20B5 Cedi Sign
U+20BA Turkish Lira Sign
U+20BD Ruble Sign
U+2117 Sound Recording Copyright
U+FB01 Latin Small Ligature Fi
U+FB02 Latin Small Ligature Fl

Google Plus

U+2052 Commercial Minus Sign
U+20A1 Colon Sign
U+20A6 Naira Sign
U+20A9 Won Sign
U+20AD Kip Sign
U+20B2 Guarani Sign
U+20B5 Cedi Sign
U+20BA Turkish Lira Sign
U+20BC Manat Sign
U+20BD Ruble Sign
U+FB01 Latin Small Ligature Fi
U+FB02 Latin Small Ligature Fl

Google Pro

U+2052 Commercial Minus Sign
U+20A1 Colon Sign
U+20A6 Naira Sign
U+20A9 Won Sign
U+20AD Kip Sign
U+20B2 Guarani Sign
U+20B5 Cedi Sign
U+20BA Turkish Lira Sign
U+20BC Manat Sign
U+20BD Ruble Sign
U+FB01 Latin Small Ligature Fi
U+FB02 Latin Small Ligature Fl

Google Expert

U+2052 Commercial Minus Sign
U+20A1 Colon Sign
U+20A6 Naira Sign
U+20A9 Won Sign
U+20AD Kip Sign
U+20B2 Guarani Sign
U+20B5 Cedi Sign
U+20BA Turkish Lira Sign
U+20BC Manat Sign
U+20BD Ruble Sign
U+FB01 Latin Small Ligature Fi
U+FB02 Latin Small Ligature Fl

Hope that helps :-)

Cheers, Ian
Libertinus Serif.txt
Libertinus Sans.txt
Libertinus Math.txt

[iandoug]

Lib Math is actually better than most. These are the files like attached above, smaller == better.

[iandoug]

I decided to get Woke and stop being so binary ....

Attached state of play for my attempt to find "best font for Western Academic Writing" based on glyph coverage rather than "just another Helvetica or Times New Roman clone".

Suggestions re group composition welcome.

Fonts included: Lib Serif, Lib Math, Lib Sans, STIX, Symbola.

It looks like Lib Serif and Math could do with some cross-pollination?

Possibly some groups need to be weighted more in the final average ... splitting Latin, non-Latin and Maths into Core and Plus already addressed that to a degree. Maybe need separate Humanities and Science averages. Of 138 fonts tested, only 12 plus the Kurinto megafonts got more than 50%.

Will modify the "Standards comparisons" code to work in similar way.

[Edit: replaced images with averages based on glyph counts not block averages]

Cheers, Ian

[iandoug]

FWIW, if I did my sums correctly, then of the current 2316 fonts in Google Fonts, we have:

(only) Google Core : 1827
Google Core + Google Plus: 297
Google Core + Google Plus + Google Pro: 71
Google Core + Google Plus + Google Pro + Google Expert: 120

But those at Expert level are different variants from only 9 families.

So Libertinus would be joining a relatively select group. (Fonts like Symbola and Quivira are not in Google Fonts)

Same applies with the Adobe standards ... there are not that many commercial typefaces at Latin 4 and above.

[Crissov]

@iandoug You might be interested in replacing Junicode in your charts by JuniusX / Junicode New, which is being developed here: https://github.com/psb1558/Junicode-New

Regarding MUFI support, Khaled Hosny once deemed it out of scope #101, but I guess that decision could be revisited if someone was willingto do the work.

[iandoug]

@Crissov
Been working on major update to that list.

I did actually look at JuniusX , but "All this is very much in progress." at the bottom stopped me from seeing they do already have files available.

Will add to my collection to evaluate. Also busy looking at the fonts on CTAN but a lot of them are not .ttf / .otf. Suppose I could convert them ...

Thanks, Ian