16:01:04 <RRSAgent> RRSAgent has joined #webfonts
16:01:04 <RRSAgent> logging to https://www.w3.org/2022/10/04-webfonts-irc
16:01:19 <Zakim> Zakim has joined #webfonts
16:01:29 <Vlad> zakim, start meeting
16:01:29 <Zakim> RRSAgent, make logs Public
16:01:30 <Zakim> Meeting: Web Fonts Working Group Teleconference
16:01:44 <Vlad> present+
16:01:51 <skef> present+
16:02:43 <Garret> present+
16:03:07 <jhudson> jhudson has joined #webfonts
16:06:56 <Garret> https://docs.google.com/document/d/1MdtB_WPC2grAx3vFgLHA1-CqRzQtWj_cJYx40W2VQgI/edit#heading=h.dy3wd9u6su7
16:07:45 <skef> Speeding up incremental transfer patch generation
16:07:58 <Vlad> scribenick: skef
16:08:16 <skef> What is the current rough cost of patch generation using the current two methods relative to subset generation?
16:09:56 <skef> garret will answer in a bit
16:10:24 <skef> Different methods of speeding up patching
16:10:27 <Vlad> present+ jhudson
16:10:42 <skef> 1) brotli quality parameter (computation vs size)
16:11:11 <skef> 2) Custom brotli encoding possibilities
16:12:11 <skef> (brotli has meta-blocks, which are semi-independent chunks of data. meta-blocks can be produced via different methods)
16:13:05 <skef> Case 1: Supporting precompression of file sections that don't change between subsets (e.g. GPOS/GSUB with fixed GIDs)
16:13:36 <skef> Can embed chunks that are pre-compressed at higher quality
16:14:58 <skef> Case ?: Custom compression and diffing of font tables with a known format
16:24:55 <skef> Data summary: "Immutable" layout slower and larger for subsetting, but patches are correspondingly smaller.
16:25:29 <skef> Patch generation times are a bit more than double subset generation times for equivalent levels of compression.
16:27:29 <skef> Noto sans data: much larger font, correspondingly longer times, but mostly proportional to the times for a smaller font
16:29:28 <skef> Brotli level 5 is the minimum for patch generation (probably because of lack of back-tracking at lower levels)
16:37:10 <skef> Some preliminary work on generating back-track maps using table-specific knowledge.
16:37:35 <skef> Also investigating the possibility of pre-compressing frequently-transmitted-together glyph subsets
16:38:15 <skef> Opportunities to speed up HarfBuzz's mode for GID preserving subset generation
16:39:34 <Garret> Data I presented: https://docs.google.com/spreadsheets/d/1Xf2KY4sbR_mSXDRJdO7mKKEBPSEBPmjCYxeqhvLOe7c/edit#gid=1341899720
16:49:40 <Garret> https://github.com/garretrieger/patch-subset-incxfer/blob/custom_brotli/util/brotli_stream.h
16:50:27 <Garret> https://github.com/garretrieger/patch-subset-incxfer/blob/custom_brotli/util/precompress-test.cc
16:51:38 <skef> Good to get comparable stats on vcdiff and also just gzipping the patches
16:53:13 <Garret> Take aways: 1. precompression looks useful, provides meaningful speedups in some cases. Also may reduce overall transfer sizes by avoiding patching layout (or other tables0.
16:53:29 <Garret> 2. Brotli quality 5-8 looks like a sweet spot for dynamic use.
16:54:07 <Garret> 3. Lots of potential in the glyf diffing and custom brotli encoder approach, more work is needed to see the full benefits.
16:54:47 <skef> Relative to potential PRs: skef has the impression that everything discussed actually fits within the current spec -- ti's all below the level of what we've recorded so far
16:58:17 <Vlad> zakim, who is here?
16:58:17 <Zakim> Present: Vlad, skef, Garret, jhudson
16:58:19 <Zakim> On IRC I see jhudson, Zakim, RRSAgent, Vlad, Garret, skef, MariaAngelaPileri, Mike5Matrix, Github, drott, emilio
16:58:38 <Vlad> rrsagent, make minutes
16:58:38 <RRSAgent> I have made the request to generate https://www.w3.org/2022/10/04-webfonts-minutes.html Vlad