WebFonts Working Group Teleconference -- 16 Apr 2014

<trackbot> Date: 16 April 2014

<scribe> scribenick: ChrisL

Vlad: we spent a lot of time to make state of the art font compression and to prove it
... serving millions of fonts to billions of users, big savings
... moderately complex code with a lot of benefits
... can have two function calls, one for woff2 and one for generic brotli

kuettel: jkew's point is a good one, more tools could support woff2
... whether that is enough to make it optional is a separate discussion but see his point
... trade off between fewer tools and better overall compression. low impact on high volume sites, more for specialized cases

Vlad: everything can be brotli compressed, but that is only part of woff2. similar to jpeg compression for images
... agree brotli has widespread use, but we want to benefit from the work we did for two years, it helps users

kuettel: all good points

Vlad: complexity of preprocessing - can be implemented in any language. only time will tell on the uptake
... only one person writes a library, gets used by everyone
... MonoType tools all written in Python and shared for servers, desktop etc
... wish jkew was on the call

ChrisL: jkew did say he was fine to move on if the group agrees

kuettel: would like to see us make a decision

raph: coming down to tradeoff between spec simplicity/few options/easy testability vs ease of simple tools

ChrisL: the simple easy script could use woff1

raph: implementing brotli from scratch in a scripting language would have poor performance, over time there will be callouts to native libraries as with gzip
... deployment time

Vlad: main consideration is effect on users, don't want to see them download more data than was needed
... code written once, deployed many times and data downloaded billions of times

raph: so point is that we don't want substandard tools deployed

Vlad: priority of constituencies
... on mobile with low cpu, also metered bandwidth so best to have small data
... do we agree to make preprocessing step mandatory and eliminate options

(agreement)

resolved: no optional parts in woff preprocessing

raph: good discussion, an informed decision

<kuettel> Actually that was Raph, but I agree :)

<kuettel> Known table tag proposal: https://docs.google.com/a/google.com/spreadsheets/d/111MT0l7LOVqotAnMXD4PMOm36jTPSznUigJPfxUYY_0/edit#gid=0

known table tags

Vlad: thanks for the detailed analysis
... tables in limbo for a long time or not frequently used need not be listed
... will not loose data or affect function, just a few more bytes

raph: 4 more

ChrisL: for new stuff (as opposed to old no longer used stuff) usage will increase over time

Vlad: some tables fell out of favour
... if widely deployed, its in. if its niche use case, can still be used
... need a policy for this, explain the choice

kuettel: caveat, limited corpus, mostly google fonts and foundry search integration. no cff fonts
... we could get more data

raph: happy with list as it stands, behedad said its fine and not to agonize over it.
... we can leave light green tags in
... these are things we see in real fonts. low wastefulness
... cost of missing a tag is trivial, number of fonts with missed tags is tiny except for future standardization
... need language for the rationale but suggest we accept this proposal

Vlad: agree
... we used 7 bits for tag code?

raph: yes, up to 7 available

Vlad: so 46 used out of 128
... include everything we know today, sill plenty room for later standardization
... even if the number doubles we can still include everything

kuettel: second tab has a lot more tables. fontlab has tons of them

Vlad: john hudson said font dev tools use that extensibiity for source tables. occasionally these slip through. we don't want to include those ones
... should never be in a final font that is an actual product

raph: spec has a guidance role. do not encourage people to ship fonts with that stuff

Vlad: publised specifications is the cut
... opentype/off, aat, graphite
... any tables in shipped fon, not the source ones

kuettel: (true type accent table, scribe missed)

raph: vtt tables not included
... but include aat ones as they are a public spec
... so its all in the first sheet plus the aat ones?

kuettel: in the second tab the non standard ones are struck out

ChrisL: we should aleways drop dsig in woff2

kuettel: not seeing many of the older truetype tables
... then graphite, a mix, rows 66 to 70 are documented but others are not
... are we planning for any sil table?
... then EPAR; after that random ones

Vlad: the crossed out ones should be disallowed. also EPAR proposal, not current
... no tools read it
... if a font does use it it is still preserved

kuettel: rows 45 to 64 are the older truetype tables

raph: there are published aat fonts although low use, not in corpus, some still around though

kuettel: row 76, SIL Silt is not in the spec
... not an expert on that

http://scripts.sil.org/cms/scripts/page.php?site_id=nrsi&id=typetuner

raph: Silt is present in the fonts but is tool specific, not aimed at end consumer, so we should not include it. should be stripped

Vlad: we need a few paragraphs to inform implementors that tool specific tables should be removed before woff2 encoding
... as we did for woff1
... separate para for DSIG to explain why its dropped, because woff2 processing invaliudates it

ChrisL: this tablular data should go in the evaluation report

kuettel: good to briefly talk about ordering

start with the microsoft OT tables, 29-30, then list in order of specification date, apple tt tables next, math, graphite, then colorfonts

Vlad: math and color are part of OFF v3
... preferto mention the ISO specs first

kuettel: so move math table up

Vlad: math and all the color ones

ChrisL: what about sbix, should include but no spec

Vlad: we can say its part of aat

kuettel: so move math up to 35 then sbix down to other aat ones

Vlad: makes sense

kuettel: row 51 the arbitrary tag follows

raph: needed for future expansion

kuettel: reserved ones?

Vlad: about half is left
... reserved for future use
... 29..31 willbe taken by identified tags unless some reason to keep the literal 31 value

raph: no its a legacy from 5 bits, arbitrary was last one

Vlad: so the arbitrary is 127

kuettel: can make a third tab with the fewer plus aat and reordering

fpwd and outreach

Vlad: spoke with PLH who mentioned AC meeting in June, team busy getting ready for it, limited resources for fpwd and outreach before meeting, much more resouce from Comm team after the meeting
... so mid June
... one considerastion if we want to have a press release and so on

raph: preference to do FPWD first and not block on publicity
... progress on solid spec is more important

Vlad: agree, just wondering because FPWD of Woff1 had a lot of PR, lots of attention
... produced a lot of feedback

kuettel: prefer to make available sooner, other ways to get more eyes on it. tweet about it etc
... there was a lot of controversy on webfonts before while now its accepted
... so we get technical review soon

Vlad: twitter and blogs has a lot more weight than official PR

ChrisL: also would prefer early publication

Vlad: so, with the approved edits, all agreed for FPWD of Woff2

(explanation of process)

RESOLUTION: Publish FPWD of WOFF2

- DRAFT -

WebFonts Working Group Teleconference

16 Apr 2014

Attendees

Contents

known table tags

fpwd and outreach

Summary of Action Items

Scribe.perl diagnostic output