Meeting minutes
Revisit FeatureTagSet encoding scheme and use of references
<Garret> Topic 1 reviewing feature tag set PR. This adds the ability to remove features from the default set. Only remaining issue is the inability for bikeshed to keep the microsoft TTF spec out of the normative section.
<Garret> Options are only reference OFF at the expense of being able to link to sections, find a work around in bikeshed, or add a post processing step that fixes up the reference.
<Garret> Decided to leave it as is for now (using TTF spec references) and pursue a fix in bikeshed.
<Garret> Going to add some additional description to the feature tag appendix that the client should select the features it needs without regard to the default set, unless it doesn't care about features in which case the default set should provide a reasonably good experience.
Using Unicode_Ideograph property to identify codepoints for obfuscation
<Garret> Skef raised concerns about possibly needing obfuscation for codepoints outside of the CJK space (eg. math or infrequently used symbols).
<Garret> Garret: noted that the obfuscation scheme isn't intended to be 100% effective as that's nearly impossible but aims to achieve a good amount of obfuscation for the average use case. So if these sort of examples are infrequently occuring in practice it's less of a concern.
<Garret> Garret: also note that we force the use of HTTPS which limits the visibility of the request to only the font server (no one in the middle can inspect it).
<Garret> Decided for now, as long as the privacy reviewer is happy with the approach we will leave as is.
Review and update open issues
https://
Shared Brotli ID expiration (https://
Possible way to solve it is to copy the relevant portion of the spec
<Garret> Garret: will check in with the compression team and figure out what their plan is.
Next call (July 18) is cancelled, we will reconvene on July 25th.