It mostly matters for cases where the glyph stream is used to reconstruct the character stream. This is mostly a corner case these days, primarily occurring when a print stream is being used to create a PDF (or the like) without access to the original font separately.
Although there are workflows where this is even normal, they are rarer than they used to be—and they were never terribly common.
So in other words, my problem is that I thought that while characters have semantics, I also understood that glyphs do not have any semantics, they're just geometrical descriptions of areas to be painted black.
Glyphs can have semantics. That’s what the cmap table does (or cmap + OpenType features). It’s the importance/meaning of that mapping that we are debating....
As Tom notes, there are circumstances in which text content can be changed based on parsing of the glyph string, although this is less common than it used to be.
I remember that, back in the early days of OpenType, Adobe registered a 'Diphthong' feature with the idea that it would ligate e.g.
a e -> æ
o e -> œ
There are all sorts of problems with that idea, and Adobe realised this and the feature was removed, and that's the point at which I recall a general principle being established that GSUB features shouldn't be used to point to glyphs of other characters.
There are some security concerns involved here, as well, as masking of underlying characters is a phishing technique.
Comments
Although there are workflows where this is even normal, they are rarer than they used to be—and they were never terribly common.
John Savard said:
Glyphs can have semantics. That’s what the cmap table does (or cmap + OpenType features). It’s the importance/meaning of that mapping that we are debating....
I remember that, back in the early days of OpenType, Adobe registered a 'Diphthong' feature with the idea that it would ligate e.g.
a e -> æ
o e -> œ
There are all sorts of problems with that idea, and Adobe realised this and the feature was removed, and that's the point at which I recall a general principle being established that GSUB features shouldn't be used to point to glyphs of other characters.
There are some security concerns involved here, as well, as masking of underlying characters is a phishing technique.