Meeting minutes
IFT client tests updates.
Scott added the ift encoder to tests and support for multiple outline formats (CFF/glyf).
Garret will review in the next couple of days.
review test plan from tpac
Skef thought of another test case. Very huge seldom used layout feature (ie. aalt).
Skef may want to add support via table keyed patch, that adds support for the feature everywhere. Then that introduces additional patches that may be needed.
Garret: tough to support for a table/glyph keyed mix since we can only rewrite the entire glyph keyed map. Can be supported in pure table keyed setup.
Skef: right, let's forget about it for now then. Not particularily important.
Vlad: do we have tests cover invalidation?
Vlad: checked the test plan and it does look like what we do. Subclause 4.3 on extension algorithm. Looks like scoped to be tested.
Garret: next steps are to convert the working to document to markdown and check into the test repo.
complex glyph substitution dicussion.
Skef: Garret's been working on a possible approach. Will need to see more details to evaluate. Have concerns.
Garret: I think the next big step is to evaluate the closure analysis across a bunch of fonts to see how big of a problem fallback glyphs are in practice. This will inform how we approach the problem (eg. use a simple but not full solution or a more complex one).
Vlad: this is probably something we want to add guidance for in the encoding considerations section.
Skef: our test plan currently does not include any mention of colour fonts.
Garret: we should definitely include that in tests.
Garret: we don't currently allow any of the colour tables to be patched by glyph keyed patch (eg. sbix), we should review and see if we want to include any.
Skef: with colour and variations you can have more extensive compositions since overlaps is allowed. So more complex dependencies. eg. lots of reused elements over a bunch of codepoints.
Garret: I think (but haven't tested yet) that emoji should be relatively well handled by the current implementation. Definitely needs more testing
Vlad: are skin tone modifiers going to be a challenge?
Garret: I don't think so, those won't typically be implemented using unique glyphs so don't affect the glyph depenency analysis. COLRv1 doesn't get glyph keyed patch.
<skef> We need to add something to the encoder section of the specification warning about shapers that compose or decompose accented glyphs (etc.) on their own.
Garret: another issue we need to think about which isn't currently handled is unicode composition and decomposition needs to be included in the dependency analysis. Shapers may automatically make compositions based on the cmap presence (which due to table keyed patching includes things which aren't fully loaded into the font yet). So glyph
dependencies need to take this into account. Ran into this during development of the demo.