28 Oct 2015

See also: IRC log


paulc, giuseppe, MikkoT, karl


<danbri> intro from cwilso

Is anyone scribing?

Alex: This came from talking to many publishers
... this is resulting in several seconds of improvements
... @@ beginning of this is unscribed.

@@@who: clarify the benefits from this modifications

Alex: It is not a modification. It is a set of web components
... Polymer, Amp
... they define their own tags
... result of extensibility
... focused on performance for publishers
... HTML Content is quite fast
... 1000 words and a video can load quick.
... problem is stuff on the outside
... font requests, script requests.
... AMP does not let you do this.

Yoav: SEO benefits

<jeff_> Alex: No active content. Helps with security. Others (e.g. Twitter can use)

<jeff_> @@@Questioner: So end user benefits

<jeff_> ... Big web sites can support more users.

<jeff_> Paul: Why did you limit this to publishers

<jeff_> Alex: Mostly because we've limited what you can do; better fit for publishers

<jeff_> scibenick: jeff

<jeff_> scribenick: jeff

<jeff_> [Alex does Amp demo with kittens]

<jeff_> Alex: Demo says - after you've downloaded please install service worker

<jeff_> ... this will result in consistent performance despite the network

<jeff_> ... On the second load we get a richer show

<jeff_> ... like an applications

<jeff_> .... responsive design

<jeff_> ... 50frames per second animation

<jeff_> ... not possible with AMP today

<jeff_> ... but consistent because the surround came from local disk using service worker

<jeff_> ... We can also extend set of elements

<danbri> demo is at https://infrequently.org/15/amp-up/demo/amp-shell-demo.html

<jeff_> ... carefully consider each element that it won't kill performance

<jeff_> ... too easy to hurt performance.

<jeff_> Chris: We call this "for publishing" because these docs are mostly static

<jeff_> ... also they see big need for performance and user experience

<jeff_> KarlD: What's in it for Google.

<jeff_> Chris: Google wants the web to be fast.

<jeff_> Karl: I mean Google.

<jeff_> DanBri: When the web sucks we suck

<jeff_> Chris: Simple things should be on a fast path.

<jeff_> ... bumper bowling

<jeff_> ... want to do more - take off the bumpers

<jeff_> ... we'll add more web components into library

<jeff_> Karl: In the demo there are links to the amp project

<jeff_> ... will these scripts be local?

<jeff_> ... if you remove the CDN of amp project

<jeff_> Chris: THere is an overestimation of what AMP is.

<jeff_> ... just a library to keep performance good.

<jeff_> ... no ISO standard for AMP

<jeff_> Yoav: Would it be AMP for SEO and app integration benefits

<jeff_> Chris: Today - no. That is driven by experience not AMP-ness

<jeff_> Alex: This is an active discussion.

<jeff_> ... I think we want speed. Ask for speed not AMP.

<jeff_> ... out of standards context - we are just beginning to address the perf problem

<jeff_> ... "AMP for Speed is good - but we want Speed"

<jeff_> Chris: THe focus is on AMP because speed is hard to define

<jeff_> ... page load time? - just one metric

<jeff_> ... hard to teach people to design for speed

<jeff_> ... we give dev tools, pattern, but not the only tactic.

<jeff_> Alex: Easy to screw-up scrolling performance, layout, resource loading, network, etc.

<jeff_> ... bumper bowling.

<jeff_> ?: Carousel?

<jeff_> Chris: Gives you carousel.

<jeff_> Alex: They are taking poll requests on github

<jeff_> Chris: Under discussion

<Zakim> tzviya, you wanted to ask for more info on publishers

<jeff_> Tzviya(DPUB IG): You mentioned publishing. Say more about publishing.

<jeff_> Alex: We mean a large news website

<jeff_> ... Do you mean dead-tree conversions?

<jeff_> Tzviya: We are looking at books - but no paper; no dead trees.

<jeff_> Alex: Ad team is working with news publishers.

<jeff_> [Chris, please drop the link for the list of publishers?]

<cwilso> It's in AMPproject.org

<jeff_> Paul: Set aside web components approach

<jeff_> ... you are effectively getting better performance by profiling HTML and the OWP

<jeff_> ... will you bring this profile to W3C?

<jeff_> ... this profile gives you good performance.

<jeff_> Alex: Last year we ripped stuff out of Blink as an experiment

<jeff_> ... e.g. no descendant selectors for CSS

<jeff_> ... [other examples]

<jeff_> ... we can do better is what we learned.

<jeff_> ... but AMP did not come from Chrome.

<jeff_> ... this came from search team

<jeff_> ... working with content producers

<jeff_> ... we'll have to bring them in.

<jeff_> ... they are talking to us [Chrome].

<jeff_> ... at that point we can bring this in as a proposal.

<jeff_> Paul. By profile I mean a constrained subset.

<jeff_> ... if AMP elements are different syntax with constrained semantics, you could apply the semantics to the original HTML element

<jeff_> Chris: In some cases we can

<jeff_> ... but not all

<jeff_> ... Embedding Twitter element is not going to be baked into HTML.

<jeff_> ... we could have a fast path and slow pather

<jeff_> ... not sure if we want to do that

<jeff_> ... can consider

<jeff_> ... this informs people how to write fast code.

<jeff_> ... may not be amp

<jeff_> Alex: This is not necessarily fast.

<jeff_> ... skip preload scanner

<jeff_> ... crazy that this ends up the general pattern.

<jeff_> ... agree we need to bring these gains into the platform.

<jeff_> ... contrafactual is that web sites are slow

<jeff_> ... this is one of many exercises to find new elements

<jeff_> ... not necessarily done in a standards body.

<jeff_> ... components allows folks to invent without us

<jeff_> ... no permission or consent

<jeff_> Dino: How does the project decide what to include?

<jeff_> ... why Twitter? why not Weibo?

<jeff_> Chris: Like any other project. Pull requests.

<jeff_> ... responsive to publishers

<jeff_> Dino: But they say "monetize" Which ad networks?

<jeff_> Alex: Publishers care about analytics packages.

<jeff_> ... working on open source versions

<danbri> dino, see also https://github.com/ampproject/amphtml/blob/master/CONTRIBUTING.md "Important criteria for acceptance of an Extended Component is that an http://oembed.com/ integration already exists and that the thing being integrated has an actively maintained Wikipedia page."

<jeff_> Dino: If I were a different browser I could choose not to load

<jeff_> Alex: Like anything else.

<jeff_> Arnaud: For the record, pull requests are requests - not how they get decided.

<jeff_> Chris: True. Goal is a fast content platform.

<danbri> also see https://github.com/ampproject/amphtml/tree/master/3p

<jeff_> ... I don't make the decisions. I can only explain their motivation.

<jeff_> Dino: Fast and with certain features.

<jeff_> Chris: Yes.

<jeff_> [everyone agreeing]

<jeff_> Chris: Powerful enough

<jeff_> ... long list of new requirements. If demand, it will happen.

<jeff_> Dino: As a browser vendor this is awesome. I don't need to do anything to benefit.

<jeff_> Alex: They'd like to not need web component polyfill.

<jeff_> Alex: Replace content. Knowing aspect ratio. User experience is fluid.

<jeff_> ... later content doesn't re-flow.

<jeff_> ... people do this with images all the time.

<jeff_> ... hard to get aspect ratio with responsive design.

<jeff_> Chris: To be clear, anything in amp can be done in HTML.

<jeff_> ... just need those bumpers.

<jeff_> Alex: CSS Replace content has many problems.

<jeff_> [discussion of different approaches to achieve same goal]

<jeff_> Yoav: Regarding enforceability, this should be for general fast web

<jeff_> ... but OTOH; can't guarantee performance from static analysis

<jeff_> Alex: A flaw in AMP is that you can make slow AMP documents

<jeff_> rrsagents, make minutes

<jeff_> ... fonts are a mess

<jeff_> ... match fonts with styles; won't fetch ahead of time

<jeff_> ... hits performance

<jeff_> ... hard to get correct

<jeff_> ... browsers want to be lazy

<jeff_> ... Chrome holds on for 3 more seconds

<jeff_> ... amp can wait for CSS enhancements

<jeff_> Yoav: Other way. If I have a fast non-AMP; how do I get twitter integration

<jeff_> ... I need to lazy load images

<jeff_> Alex: Yup. Bug in the way the content is consumed.

<jeff_> ... we need an upgrade strategy.

<jeff_> Chris: We define fast irrespective of amp

<jeff_> ... webaudio for SEO

<jeff_> Alex: You don't need polyfills.

<jeff_> Yoav: But if I want it to work w/o customer elements...

<cwilso> S/we define/we should define/

<jeff_> Alex: Not ideal. Fudges

<jeff_> ... still Big for publishers

<jeff_> Yoav: Iterating in user land, but can we do better?

<jeff_> Alex: Yes.

<jeff_> Yoav: Bake into platform?

<jeff_> ... browser can enforce.

<jeff_> Alex: Let's talk!

<jeff_> Alex: Shall we talk about breaking the web?

<jeff_> Karl: You talked to a lot of publishers.

<jeff_> ... why did you need this solution.

<jeff_> ... tools exist to improve performance.

<jeff_> Alex: Without something that changes relationship between business and dev can't fix the problem

<jeff_> ... even Ilya can't get it done.

<jeff_> Ilya: Alex is right.

<jeff_> ... can't remove tags

<jeff_> ... some other constraint.

<jeff_> MikeChamp: How does Amp solve?

<jeff_> Alex: Handwaving to business. This is how it is done!

<jeff_> Mike: How does it help?

<jeff_> Ilya: It's a reset>

<jeff_> Jeff: So make it a standard - strengthens the handwaving

<jeff_> Alex: Validator is key

<jeff_> ... combined with UI treatment, developer story

<jeff_> ... so yes, maybe.

<jeff_> ... I think of it as a way to evolve HTML.

<jeff_> ... since they are custom elements - a crawler will tell us what people are using.

<jeff_> ... Hixie did it for HTML5

<jeff_> ... easier with custom elements

<jeff_> ... they have UI

<jeff_> ... they have meanings

<jeff_> ... AMP one vocab that can help

<jeff_> Chris: Just making it a standard is a challenge

<jeff_> ... this is very targeted

<jeff_> ... may need a different set for a different people.

<jeff_> Jeff: That's why Paul was calling it a profile.

<jeff_> Chris: we can get them, but it is super-early.

<jeff_> ... not yet published

<jeff_> Ilya: Plenty of case studies (hundreds)

<jeff_> ... but not compelling

<cwilso> S/can/will/

<jeff_> ... not optimizing speed.

<jeff_> ... AMP changes the equationg

<jeff_> ... you had been doing it wrong

<jeff_> @@: That's a measurement problem.

<jeff_> Ilya: Still not compelling enough.

<jeff_> Paul: Programmers are infected to do things complicated

<jeff_> ... hard to get them to simplify

<jeff_> ... engineering/business conflict

<jeff_> Arnaud: amp-html tutorial says it is for publishing content everywhere

<jeff_> ... you might tone it down

<jeff_> ... create-page md

<jeff_> Alex: Send a pull request

<jeff_> Arnaud: You got it --- Verbally

<jeff_> [Google people scramble to find it]

<Arnaud> freudian slip?

<jeff_> Yoav: If desire is a fast-mode; what's the plan to get publishers there.

<jeff_> Chris: Plan is to identify what fast means.

<jeff_> ... get people to be fast.

<jeff_> ... I don't care how they do it.

<jeff_> ... I won't penalize them if they use angular

<jeff_> Yoav: Assuming you can identify

<jeff_> Chris: Difficult challenge

<jeff_> [ovation}

<jeff_> [adjourned]

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2015/10/28 05:28:57 $

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 http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Found ScribeNick: jeff
WARNING: No scribe lines found matching ScribeNick pattern: <jeff> ...
Inferring Scribes: jeff

WARNING: No "Topic:" lines found.

Present: paulc giuseppe MikkoT karl

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

Got date from IRC log name: 28 Oct 2015
Guessing minutes URL: http://www.w3.org/2015/10/28-amp-minutes.html
People with action items: 

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]