Meeting minutes
To summarize a discussion that happened at the start of the meeting:
skef is not currently as optimistic about the eventual use of the static extensions to patch-subset as some other members (given the minutes of the recent TPAC meeting)
One reason is that the granularity of subsets when using the static conventions is much more coarse than it is for the other technologies we have discussed. And the mitigation strategies for this limitation (mixing in top-level static subsetting to reduce the scope of the static-brotli patches, taking more specific expected use cases (e.g. which languages will usually be used) add their own complexity)
Also, when mixing an IFTB-like system with patch-subset, e.g. for allowing selection and augmentation of variable axes, you have similar explosion problems: 10 axes leads to 1024 sets of IFTB-like chunks, and this assumes no augmetation (just replacement) of chunk data when changing axis counts. (If you wanted to patch chunks, if that even works, you would need 9! other chunk sets)
<Garret> Garret: this potential new thing is still very conceptual and definitely needs more exploration and validation before we officially decide this is the path we want to take. So for next steps I want to spend some time and develop a concrete proposal for an updated mapping table which will be able to support dependent (shared brotli) and independent patches (IFTB) and also likely a small prototype to help assess viability.
<Garret> Garret: In the mean time we should continue to make progress on the IFTB spec PR.
<Garret> Garret: I'm currently skeptical about the viability of mixing dependent and independent patches in the same font. I believe it can be done, but unsure that the additional complexity is worth the benefits it will give. However, I think this is still worth some more research.