W3C

- DRAFT -

Web Performance Working Group

26 Jul 2017

See also: IRC log

Attendees

Present
igrigorik, shubhie, Nic, todd, xiaoqian, nolanlawson
Regrets
Chair
igrigorik
Scribe
nolanlawson

Contents


Spec Audit

<igrigorik> spec audit: https://docs.google.com/document/d/106fu-2acn7JexSadxyKrxQny78WROk6u-Vwlte2zpgg/edit

okay taking notes now, please correct me if I goof something

ilya: rIC is ready to go to CR
... blocking issue with Beacon is tests

toddreif: we're trying to focus on finishing the PR
... we have to delete worker tests because that was removed from the spec. tests need minor cleanup

ilya: for HR-time need tests for time-origin

toddreif: should be done by EOW

ilya: should we merge time-origin into L2? it's in firefox and soon in chrome

toddreif: it's in Edge insider too
... we'll see if we can take on tests

ilya: ok so it'll go in L2
... need to revisit performance timeline before tpac

toddreif: should we go back to working draft?

ilya: we need to manually push a CR update every time
... we'll go back to working draft and revisit after TPAC
... landed tests for page visibility. missing is onvisibilitychange handler. safari supports, firefox implementing. merge into L2?

preload: many shipping impl's. let's push for CR by TPAC. problem is unspecced cache behavior

(ilya^)

ilya: long tasks has editorial issues. is v2 L1?

shubhie: haven't done any work on v2 yet, maybe not worth blocking

toddreif: we'd prefer to skip v1 and ship v2

shubhie: firefox is actively looking at it

toddreif: maybe good to get an editorial draft. can't get to CR without 2 impls
... let's get v1 in editorial draft, then put likely v2 chrome stuff into that v1, then drive consensus

ilya: whatever firefox chooses from v1/v2 may become the L1
... let's tackle editorial issues then watch firefox/edge progress. keep as working draft
... user timing missing worker tests

toddreif: we'll push on that

ilya: server timing making progress, still in draft

Improve Links of the WebPerf Specs

toddreif: we need to ensure good links in the specs

ilya: marcos suggested better format than markdown. we have lots of small specs that are tightly coupled, links are hard to maintain. combine specs?

toddreif: performance timeline could theoretically be one spec. problem is keeping things well-defined

ilya: hr time and timeline are tightly coupled. hr time was for perf.now()

toddreif: different UAs still ship different specs in different rhythms. html5 is a mega-spec though
... let's get consistent first and see if separate specs can work
... can't have broken links

nic: people reference subsections and then sometimes those get renamed

xiaoqian: one approach is to update upstream spec. or file issue on other spec

ilya: respec vs bikeshed: what is the good pattern

xiaoqian: error reporting in other spec may help

ilya: knowing all upstream links is a hard problem

xiaoqian: one big spec doesn't help, most broken links are in html5 spec

ilya: marcos says respec has ability to pull in reference. can flag issues for unknown reference

xiaoqian: respec always keeps link to latest

PageVisibility #29

ilya: pagevisibility issue 29, can we log if page was ever visible

<igrigorik> https://github.com/w3c/page-visibility/issues/29

shubhie: devs need insight into visibility status early on, especially given background throttling. need to know before script registered

ilya: things like ctrl-click skew telemetry due to throttling

shubhie: devs want to know if page was loaded in background or foreground and whether it was ever visible/hidden

ilya: implicit here is that background === throttling. should we expose throttling itself?

shubhie: different thing, devs want to know if users got bored and tabbed away

nic: telemetry-wise both are useful

ilya: we had prerender switch, prerender-to-visible. nobody implemented. could log visibility event to cover tabbing in and out

nic: we do that as soon as we're on the page

shubhie: seems a common problem, need analytics before script can load

ilya: came up at f2f, issue is analytics can load at any point. onload is well defined point. every entry type can define its own buffer

nic: ryosuke pushed back, said max would always be full

ilya: can be worked around via loader snippets
... does surfacing new entries for page visibility make sense for L2?

toddreif: folks I've talked to just want a boolean to say if it was ever non-visible
... goal is to separate pristine foreground pages

nic: I think we want "was ever visible" too

shubhie: that was the original bug

ilya: at which point does this get flipped

toddreif: for same tab, when the document is created
... visible/hidden are separate from painted/not painted

ilya: facebook wants to know if pages loaded slowly because user switched away and page was throttled. boolean wouldn't address that problem

toddreif: pages users never touched, they want a bucket for that

shubhie: let's ask if two bits - ever hidden, ever visible - and see if that solves the issue

ilya: let's look at high resolution time as primitive PR from tim

Summary of Action Items

Summary of Resolutions

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.152 (CVS log)
$Date: 2017/07/26 18:12:01 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.152  of Date: 2017/02/06 11:04:15  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00)


WARNING: No "Present: ... " found!
You can indicate people for the Present list like this:
        <dbooth> Present: dbooth jonathan mary
        <dbooth> Present+ amy
        <amy> Present+

No ScribeNick specified.  Guessing ScribeNick: nolanlawson
Inferring Scribes: nolanlawson

WARNING: No "Topic:" lines found.


WARNING: No meeting title found!
You should specify the meeting title like this:
<dbooth> Meeting: Weekly Baking Club Meeting


WARNING: No meeting chair found!
You should specify the meeting chair like this:
<dbooth> Chair: dbooth

Got date from IRC log name: 26 Jul 2017
Guessing minutes URL: http://www.w3.org/2017/07/26-webperf-minutes.html
People with action items: 

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.


WARNING: No "Topic: ..." lines found!  
Resulting HTML may have an empty (invalid) <ol>...</ol>.

Explanation: "Topic: ..." lines are used to indicate the start of 
new discussion topics or agenda items, such as:
<dbooth> Topic: Review of Amy's report


[End of scribe.perl diagnostic output]