Meeting minutes
Updated IFT Spec
Introduction to the new draft of the specification
<Garret> https://
The specification is based on the javascript demonstration of the prototype
Comparison is to a font broken up into subsets with unicode-range
End goal: only pull parts of font file over the network that are relevant to rendering it
Unit of transfer: patches
To make this work: We add a new table to the font that indicates what patches provide what information and where they are (i.e. their urls)
Two kinds of patch: a non-semantic Brotli patch, and a semantic patch for inserting sets of glyphs into the original font subset
Question: What is the subsetter in use currently?
The demo uses the HarfBuzz subsetter to generate static files
To generate the per-glyph patches: different subsetter working differently
The subsetter itself is not part of the specification; there could be different implementations
It's possible that the per-glyph patches could be generated dynamically -- would need to build a new encoder to evaluate that
Could make data-types section a subsection introduced later
Are the new tables standardized through Open Font Format or will they be "private"?
We assume they will be private
This draft starts out imposing minimal requirements, we will evaluate what additional requirements to add going into the future.
Closure requirement left out of this draft
There is some disagreement about whether the closure requirement should be added to the spec as a "should" or a "must" - skef considers the requirement intrinsic to the per-glyph format
Chris: Should http be required?
Skef: Current URLs are relative rather than absolute
<Garret> Skef: Section 7 steps 6 and 7 unclear what to do with independent patches.
(Garret clarifies and will evaluate whether changes are needed for clarity)
Garret anticipates that we could have a relatively complete draft by March 2024
Need to add patch table format 2
May need to add restriction on how fine grained codepoint patches can be
<Garret> Garret: would be good to have an additional section that explains how the mechanism works before getting into the fine details. Skef has feedback that jumping right into section 5 is jarring.
<Garret> Garret: don't use "full font" in offline usage, this isn't accurate. What we really mean is the same font a non-offline client would have gotten.
<Garret> Skef: "only the data" in font subset. Only is too hard.
<Vlad> The next WG Zoom meeting will be held on Feb. 27.