- From: Roderick Sheeter <rsheeter@google.com>
- Date: Wed, 20 May 2015 14:05:53 -0700
- To: WebFonts WG <public-webfonts-wg@w3.org>
- Message-ID: <CABscrrE3_p=B48KArN5YLQ3yij_z+GCOurtvqBfWk03=1cDwgg@mail.gmail.com>
http://www.w3.org/2015/05/20-webfonts-minutes.html
[1]W3C
[1] http://www.w3.org/
- DRAFT -
WebFonts Working Group Teleconference
20 May 2015
See also: [2]IRC log
[2] http://www.w3.org/2015/05/20-webfonts-irc
Attendees
Present
+1.408.927.aaaa, +1.650.214.aabb, [Microsoft], Vlad,
jfkthame, +1.650.352.aacc, Ken_Lunde__Adobe_,
kuettel+RSHeeter, Behdad, +1.650.253.aadd,
+1.650.253.aaee
Regrets
John_Hudson, Chris_Lilley
Chair
SV_MEETING_CHAIR
Scribe
RSheeter
Contents
* [3]Topics
1. [4]action items
2. [5]hmtx preprocessing
3. [6]Action Items
4. [7]CTS
* [8]Summary of Action Items
__________________________________________________________
<trackbot> Date: 20 May 2015
Khaled, did you get an invite to a hangout on your gmail?
action items
<Vlad> [9]https://www.w3.org/Fonts/WG/track/actions/open
[9] https://www.w3.org/Fonts/WG/track/actions/open
<Vlad> New topic: CFF support in WOFF2
<scribe> scribenick: RSheeter
(discussion of ability to optimize cff)
Vlad: if de-subroutinization allows better compression with
brotli, would we have to re-sub on receiving end Jungshik: no,
but memory use might be greater. Perhaps faster to use as well.
... summary, some parts can be just dumped during encoding to
save bytes. Less constraints here (server/fast wkstn).
... impact on decoder is of greater concern. Eg the memory use
cost in return for smaller compressed file.
Ken: concur we should not re-subroutinize when decoding
... could be ~10% increase in filesize without re-sub. Also
re-sub is *very* slow [potentially hours]. And lots of memory
used.
(looking for link to doc about gains from de-sub)
<Vlad>
[10]https://docs.google.com/spreadsheets/d/1jvJ3xUOg6rySiAtiXDv
Af7Du43XytEBFJCr5YO12F7c/edit#gid=0
[10] https://docs.google.com/spreadsheets/d/1jvJ3xUOg6rySiAtiXDvAf7Du43XytEBFJCr5YO12F7c/edit#gid=0
Vlad: So 5-10% filesize reduction in output woff2 if we desub.
Is it worth adding as a mandatory step? (at cost of larger
decompressed file)
Behdad: why Mandatory?
jfkthame: we shouldn't make this mandatory
<Ken_Lunde__Adobe_> I agree that desubroutinized CFFs should
not be mandatory.
Vlad: agree; doesn't have to be normative. Maybe add a note to
spec that describes tradeoffs of doing this but don't
recommend; just informative.
Behdad: just note it in the spec
action to vlad Add description of tradeoffs in cff
preprocessing (desub, etc)
<trackbot> Error finding 'to'. You can review and register
nicknames at <[11]http://www.w3.org/Fonts/WG/track/users>.
[11] http://www.w3.org/Fonts/WG/track/users
action vlad Add description of tradeoffs in cff preprocessing
(desub, etc)
<trackbot> Created ACTION-175 - Add description of tradeoffs in
cff preprocessing (desub, etc) [on Vladimir Levantovsky - due
2015-05-27].
Welcome mystery guest!
<cosimolupo> sorry the sip phone is not working.. didn't mean
to disturb, please go on.
hmtx preprocessing
Behdad: if lsb is same as glyf we can drop it and save bytes.
Easy to rebuild hmtx back from glyf. Low complexity.
Vlad: won't somebody think of the swashes?!
jfkthame: negative lsb would just be reflected in x-min of bbox
Behdad: lsb allows you to move glyf without moving points.
Usually lsb == x-min. If the same for all glyfs we can just
drop it. Estimated 90%+ use same value.
... That is, of 10 fonts 9 will have *all* lsb == x-min.
Vlad: seems easy; just need to know how to tell decoder it was
dropped.
(clarification that there is no per-table transform marker but
we do have reserved bits that could be used for flags)
Behdad: maybe we should test run htmx mangling over a test
corpus
Vlad: Can we test with Google Fonts? [yes]
action rsheeter Test hmtx transformation over Google Fonts
corpus (how many lsb == x-min for all glyfs, what savings)
<trackbot> Created ACTION-176 - Test hmtx transformation over
google fonts corpus (how many lsb == x-min for all glyfs, what
savings) [on Roderick Sheeter - due 2015-05-27].
Action Items
<Vlad> [12]https://www.w3.org/Fonts/WG/track/actions/open
[12] https://www.w3.org/Fonts/WG/track/actions/open
Vlad: Did you read my diff?
close action-173
<trackbot> Closed action-173.
action-116
<trackbot> action-116 -- David Kuettel to Decoder performance
analysis on mobile devices -- due 2015-04-22 -- OPEN
<trackbot> [13]http://www.w3.org/Fonts/WG/track/actions/116
[13] http://www.w3.org/Fonts/WG/track/actions/116
Vlad: summarize your findings
Kuettel: Chrome instrumented. Got data. Compression team hard
at work improving numbers. Caveat: environment not entirely
controlled (wild users). WOFF1 could be optimistic in some
cases.
[we may have lost audio]
unfortunately we have now disconnected Khaled :(
Vlad: can we measure how smaller file vs slower decode might
play out (eg are we ahead)?
Kuettel: will ping Kenji
(exploration of further testing, ways to get overall latency
impact)
action-148
<trackbot> action-148 -- Vladimir Levantovsky to Review and
edit the conformance test for extraneous data -- due 2015-05-06
-- OPEN
<trackbot> [14]http://www.w3.org/Fonts/WG/track/actions/148
[14] http://www.w3.org/Fonts/WG/track/actions/148
action-168
<trackbot> action-168 -- Roderick Sheeter to Measure RAM usage
for woff2 vs. woff1 -- due 2015-05-13 -- OPEN
<trackbot> [15]http://www.w3.org/Fonts/WG/track/actions/168
[15] http://www.w3.org/Fonts/WG/track/actions/168
Vlad: Everyone please review all normative statements to ensure
they are testable
Victory declared over action items
CTS
[more or less out of time]
Vlad: Ken want to come to Boston?
Ken: bit difficult
[next f2f after Boston will be ATypI. Come one, come all]
Ken: Maybe David Lemon could come to atypi f2f
Summary of Action Items
[End of minutes]
__________________________________________________________
Minutes formatted by David Booth's [16]scribe.perl version
1.140 ([17]CVS log)
$Date: 2015/05/20 21:03:53 $
__________________________________________________________
[16] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
[17] http://dev.w3.org/cvsweb/2002/scribe/
Scribe.perl diagnostic output
[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.140 of Date: 2014-11-06 18:16:30
Check for newer version at [18]http://dev.w3.org/cvsweb/~checkout~/2002/
scribe/
[18] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/
Guessing input format: RRSAgent_Text_Format (score 1.00)
Found ScribeNick: RSheeter
Inferring Scribes: RSheeter
Default Present: +1.408.927.aaaa, +1.650.214.aabb, [Microsoft], Vlad, jf
kthame, +1.650.352.aacc, Ken_Lunde__Adobe_, kuettel+RSHeeter, Behdad, +1
.650.253.aadd, +1.650.253.aaee
Present: +1.408.927.aaaa +1.650.214.aabb [Microsoft] Vlad jfkthame +1.65
0.352.aacc Ken_Lunde__Adobe_ kuettel+RSHeeter Behdad +1.650.253.aadd +1.
650.253.aaee
Regrets: John_Hudson Chris_Lilley
WARNING: No meeting chair found!
You should specify the meeting chair like this:
<dbooth> Chair: dbooth
Found Date: 20 May 2015
Guessing minutes URL: [19]http://www.w3.org/2015/05/20-webfonts-minutes.
html
People with action items:
[19] http://www.w3.org/2015/05/20-webfonts-minutes.html
WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.
[End of [20]scribe.perl diagnostic output]
[20] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
Received on Wednesday, 20 May 2015 21:06:21 UTC