See also: IRC log
<wilhelm> Scribe: wilhelm
sstewart6: I've started adding URI endpoints to the spec. That should make it easier to write conforming implementations.
Automate_: I've been trying to
make the spec look like a spec.
... User interactions stawman discussed in June is landed.
... Please raise bugs on this.
Automate_: Going through the bugs, closing some.
sstewart6: We've moving towards
... Selenium is the example implementation of this.
... Selenium 3 will be W3C compatible.
... It has not shipped yet.
lukeis: Frame searching bug?
Automate_: There is a section on bugs.
<lukeis> *switching (not searching)
sstewart6: Come up with a better name of interactable?
Automate_: No conclusion.
ato__: Suggested reachable.
Automate_: Minutes from last time cover the consensus.
sstewart6: We had a discussion with P&F WG last meeting.
Automate_: We spoke to the
webperf group last time, re: page visibility and element
... sstewart6 showed them how we'd want it to work.
... That was going to be in a future version of the spec.
... They just pushed one of their specs to CR.
... Mozilla has raised bugs on this.
sstewart6: We did rename section
... No progress on "Request elementfrompoint with a DOM relative coordinate instead of a viewport relative coordinate"
sstewart6: We might need to handle the visibility check ourselves.
MikeSmith: Has anyone said they want to work on this?
sstewart6: The closes we get is the webperf group. Use case is advertisers.
MikeSmith: Why did CSS not decide to do it?
Automate_: I raised a bug against
... I suspect it's lost along the way.
sstewart6: Discussed in 2012. CSS passed buck to DOM land, and the other way around.
MikeSmith: Make we could motivate Moz CSS rep to work on this...
<scribe> ACTION: sstewart6 to talk to CSS WG about visibility detection [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action01]
<scribe> ACTION: Automate_ to speak to Tantek about his old action item on visibility and hit testing [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action02]
Automate_: There is no good definition of hit testing anywhere.
JohnJansen: Covered in CSS OM View spec?
<MikeSmith> CSSOM View Module
Automate_: Nobody wants to define this, as it's a big can of worms.
<JohnJansen> and the more generic CSSOM spec: http://www.w3.org/TR/cssom/
<MikeSmith> CSS Object Model (CSSOM)
sstewart6: Add scrollwheel support: no progress.
Automate_: First action on me is
... No progress on upstream of screenshot tests. Has not been high priority.
... Strawman of 17.2 is now in place.
MikeSmith: No progress on link
from TR to editors draft.
... TR will point to editors draft.
... A recurring problem is that implementors use the outdated TR version of the spec, while they should be using the editors draft.
sstewart6: Some progress on
identifying tests to move over from the open source
... Placeholders have been added.
... No progress on WCAG analysis.
... Strawman for HTTP normalisation for the JSON Wire Protocol has been added to the spec.
lukeis: Is this section G?
sstewart6: That section will be automated based on content elsewhere in the spec.
Automate_: The notes in the table - how would they be different from describing how the method is working?
sstewart6: The note will make it over to section G.
Automate_: We need to conscious that different parts of the spec will go out of sync.
sstewart6: Should be a reference
to a particular section.
... We could make the notes non-normative.
sstewart6: These are the URLs that all implementations will support.
Automate_: Moz will support
... ato has done work on this recently.
... web-platform-test will make it into the main Moz tree.
... For selendroid there's going to be interesting problem you need to solve.
sstewart6: Main problem is instanziating the browser...
lukeis: I will submit patches..
sstewart6: Current URLs have a leading slash. They may not be in the root.
lukeis: We could have a prefix?
<scribe> ACTION: sstewart6 to add prefix to JSON URLs [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action03]
sstewart6: Where are we?
... We're not tracking as fast as we should.
AutomatedTester: Spec is, test
suite is not.
... Bugs have been raised on the spec
<AutomatedTester> Ms2ger: :)
wilhelm: I reported some bugs on wooly language of the spec.
AutomatedTester: Ms2ger will go
through the spec and rip it apart. He has already given useful
... (Cites examples.)
... We have a lot of fluffy language. Describing features, but lacking in confromance requirements.
MikeSmith: Input from jgraham would also be useful here.
ato__: Progress on the test suite
may be slow due to style differences in test suite.
... Using the Python package makes it harder to test lower level things.
... Capabilities are not exposed.
sstewart6: I'd use a raw HTTP
... Set aside time to discuss the style of the suite.
<ato> Ms2ger: We'll do that.
ato__: We should switch to wptserve.
<scribe> ACTION: ato__ to switch to wptserve in webdriver test suite [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action04]
<ato> wilhelm: Hi, this is me.
sstewart6: We're sort of on track with the spec, but not with the suite.
AutomatedTester: There will be a lot of changes needed to the spec, to be discovered as we work on the tests.
MikeSmith: You don't need to do
anything! I need to get off my ass and do something.
... The charter needs some more detail on milestones.
... There is a heartbeat requirement about pushing to TR.
... We'll ask for a two year extension for the charter.
jimevans1: The use case is:
suppose you have an element that is so large that even
scrolling it into the viewport, the center will not be in the
... The script won't be able to click the center.
MarcFisher: What if the center is otherwise obscured?
sstewart6: If the center is obscured: don't use click. Use interactions API.
jimevans1: Proposed: Click could still work by clicking on the midpoint of the visible area.
mdas: Why can't we just scroll to where we want to click?
jimevans1: We were lazy...
sstewart6: (Describes approach similar to what is proposed in the bug.)
(Conclusion to this issue is added to the bug.)
sstewart6: Use case: run test
with a constrained viewport. Want to compare screenshots.
... Better: Capture the top-level browsing context.
... Or the current browsing context.
... Chrome team was not happy with the whole DOM.
Shuotao_chris_: Lacking low-level tools for this...
sstewart6: If it's in the
... Stitching is an awful approach.
MarcFisher: Don't let the limitations of Chrome limit us here.
sstewart6: We resize IE windows to the document size in the IEdriver.
JohnJansen: Most of the time you
don't need to worry about what's outside the viewport.
... Use a flag?
... It's better to target a specific area.
sstewart6: Element capture.
... MS and Google is saying taking a screenshot of the entire page is difficult.
AutomatedTester: Moz implementation is a hack...
sstewart6: What is a good way to
do this? I'd like to hear from the implementors.
... Visible viewport?
JohnJansen: We'd like to have an option. Full viewport or the whole document.
sstewart6: All implementations must be able to capture the viewport. Capability for whole document.
ato: Why not just use element capture of the root element?
sstewart6: It will depend on the browser. You can't just assume it works.
AutomatedTester: If someone wanted to test their responsive web site, we don't actually support that.
sstewart6: When capturing the top
level browsing context, maintain page width for this use
... In level 2, we can describe the more complex case of rotation, etc.
... All browsers must be able to capture the viewport. All browsers should be able to capture the entire document as if the window was resized vertically. Capability for the latter.
seva: How would you request either of these?
... We should prefer the full document.
JohnJansen: Chrome disagrees with making the full DOM default. If running tests against my site, I'd need to store images for both variants.
MarcFisher: Screenshotting just the element is more sensible.
JohnJansen: Use case of position:fixed;
sstewart6: We could add a flag on the wire to indicate size.
MarcFisher: It's a POST command. We can add values to it.
sstewart6: Capability on whether
you can do these things.
... Pass a parameter.
... Default is as big as you can go.
ato: We should have a default.
seva: Interop is the argument here. If all browsers support viewport, we should default to that.
sstewart6: If you capture an
element and the element is bigger than the viewport...
... What do we do?
mdas: Use a separate canvas?
Shuotao_chris_: Won't work in ChromeDriver.
<ato> JohnJansen: Fixed
Shuotao_chris_: We don't have the low level support.
sstewart6: Fundamental constraint?
Shuotao_chris_: GPU team says no. They don't want to do a hack.
sstewart6: Shove the element into
the top left corner and capture as much as you can.
... Leave it undefined?
... We could put a note in.
JohnJansen: (Quotes spec on
... Two floats, side by side. In Chrome, there's allowed to be a overlap. In IE, no. Interop issue. We need to be able to catch that.
... For that bug, I know I need the parent element. Other times, I don't know.
... I prefer if it doesn't scroll.
sstewart6: Use case: long page, element is out of view.
JohnJansen: I'd rather scroll to view first.
sstewart6: There is no such
... Part of the audience is people without programming experience. We need to make this simple.
JohnJansen: For reftests, we could add "take screenshot without scrolling".
wilhelm: We need to support both use cases here. Reftests and magic scrolling to element.
sstewart6: Option of passing a parameter rather than a capability.
AutomatedTester: If parameter is not there, scroll. If it is there, don't.
<scribe> ACTION: AutomatedTester to add a flag to take element screenshot without scrolling [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action05]
sstewart6: Which option should it be?
<scribe> ACTION: sstewart6 to define that cookies can only be set against domains that are currently loaded [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action06]
lukeis: Use case is: set cookie before navigating to the site.
sstewart6: Don't do that.
lukeis: It should throw an exception.
sstewart6: I can add language for
... People set cookies to avoid login.
... They should navigate to a page, set a cookie, then reload.
... This is to allow a pure JS implementation of this.
ato: Greg had an opinion on this in June.
sstewart6: Use case in Maps.
MarcFisher: What interaction would you want to do with the container?
sstewart6: We need Greg to clarify.
sstewart6: Punt to level 2.
jimevans1: No, just no?
MarcFisher: Definitely level 2.
sstewart6: I have no idea how to implement this.
sstewart6: I don't think this is
... The click on element is do what I mean.
... Do as I mean APIs could be implemented in top of the interaction APIs.
... Bleedthrough expected.
sstewart6: What other data would
you like to return?
... FF does a good job of this. IE returns name and value.
JohnJansen: We could return if it's third party, session cookie, secure cookie...
<scribe> ACTION: AutomatedTester to add whether it's a session cookie, a secure cookie [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action07]
<scribe> ACTION: AutomatedTester to link to the cookies RFC to describe the data format for expiry [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action08]
ato: We reached a conclusion in Boston, but spec is not updated.
AutomatedTester: Do we want to spec files going over the wire?
<scribe> ACTION: sstewart6 to describe how we send files across the wire [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action09]
ato: This is in level 2.
ato: Since we're adding the argument to take screenshot, we'll need this.
sstewart6: Spec says you can pass whatever across the wire.
sstewart6: I'm in favour of this.
JohnJansen: That' fine.
sstewart6: Do we want to go lowercase?
<scribe> ACTION: sstewart6 to lowercase platform names [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action10]
sstewart6: That's legit.
... I'll soften the language to SHOULD.
sstewart6: This should be in
... This MUST be in level 1.
sstewart6: I'll take that.
... INteresting case of quit mulitple times.
lukeis: Case of close last window.
AutomatedTester: Language is fluffy, but says you should close...
sstewart6: We could drop id. This
is possible with CSS selectors.
... We could also drop the capabilities.
... The language bindings can still support id. But the remote end won't need to.
seva: This may confuse other
... Does the spec cover language bindings?
lukeis: For native language bindings, id is still valid.
sstewart6: We will remove id, not implement classname. We will remove capability for CSS selector.
<scribe> ACTION: sstewart6 to remove id location strategy [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action11]
<scribe> ACTION: sstewart6 to remove CSS selector capability (it's assumed) [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action12]
MarcFisher: Add note that those issues can be handled on the local end?
<scribe> ACTION: sstewart6 to add a note that you can use extension mechanisms to add location strategies [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action13]
JohnJansen: We don't support
... Must we have a must for XPath?
sstewart6: XPath is commonly used
by test authors.
... There is a JS library for this.
... All it needs to do is to appear to support it. Translating to querySelectorAll is fine.
JohnJansen: Yes, we can translate it on our end.
MikeSmith: Why do we keep XPath?
sstewart6: It's popular.
jimevans1: It's the only way to select things via visible text.
sstewart6: It's popular among QA.
wilhelm: Coud we subset it? So we don't need all of XPath?
ato: That's not a bad idea.
sstewart6: The problem is there are so many existing tests that rely on this.
JohnJansen: Should instead of must?
sstewart6: I don't want to break existing tests.
jimevans1: Traversal use cases are also not overed by selectors.
ato: Could we keep this in the language bindings?
sstewart6: Or keep it as is. More
efficient for the mobile use case.
... I hate it, but we need to keep XPath.
mdas_: Deprecation warning?
MikeSmith: Little impact.
wilhelm: Can we start the long march towards removing the use cases for XPath by subsetting?
sstewart6: Yes, but not in level
... XPath must be supported, but not neccessarily in the browser itself.
JohnJansen: Our driver won't
support XPath natively.
... We'll have to write a translator.
... We won't write our tests using XPath.
sstewart6: We put this as a must. We ask Google, Salesforce, Facebook about the prevalence of XPath.
ato: What does wpt do here?
JohnJansen: We're doing querySelectorAll for wpt.
lukeis: We have to look through RC code...
JohnJansen: Interesting data point: How many can be transslated to querySelectorAll?
sstewart6: If your translator is
avaible, people could run their tests to see what
... We'll revisit this issue when we have data.
seva: We will not define select by text?
sstewart6: Not in level 1.
<JohnJansen> So there's no confusion: The IE team is writing a web driver implementation for IE11+
sstewart6: It's a must, really. Even if it has no size, it looks like the body element is there.
lukeis: I'm okay with saying the body is always visible.
<scribe> ACTION: ato to research root element behaviour in HTML and XML [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action14]
<MikeSmith> scribe: MikeSmith
switchToFrame should only accept a WebElement or number argument
sstewart6: why should it not?
Automate_: everybody except IE doesn't follow the spec
MarcFisher: or we could just get rid of the index
Automate_: see that bug
AutomatedTester: discussion among Hixie and bz
sstewart6: what's wrong with us retaining the number and equatin
[discussion of how to address the issue]
sstewart6: if I'm a developer and I want to switch to frame at index 1, how do I implement that?
mdas: can't you get the frame element
jimevans: not across domains
AutomatedTester: can see it but can't see what's inside it
sstewart6: spec currently just says, whatever frame would be given back?
MarcFisher: but that may not
produce the same frame across browsers
... the problem is that the current behavior is not useful
sstewart6: it may be sufficiently
... question is whether this is our problem to solve
ato: we could defer this to the next f2f
sstewart6: for us I don't think
... remove the locaters by ID and name
... keep index and webElement
... and add a note that webElement is the preferred for the reason that index won't give you the same frame across browser engines
lukeis: [want to suggest we discuss problem of switching from a frame to its parent frame]
sstewart6: what do people think about switching to parent? quick vote
[everybody likes it]
sstewart6: ok we'll do it then
[agreement about speccing it by adding descriptive string (instead of say, -1)]
sstewart6: pluse new endpoint
JohnJansen: one other thing about frames
<scribe> ACTION: Luke to identify new endpoint for hooking switch to parent into [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action15]
<scribe> ACTION: Simon to define frame in section 6.1 [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action16]
Drop reference to HTML4 from the "Interactable elements" chapter
<scribe> ACTION: David to review references to ensure they meet transition requirements [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action17]
"Page load strategies" constants are ambiguous
sstewart6: I will fix that
HTTPS requirements are inconsistent
sstewart6: I will fixe that
Click does not mention if it should/shouldnt wait for page loads if its noticed (
sstewart6: that's deliberate
[discussing how to identify use case of how to log what command caused a failure]
[noting that logging still needs to be written]
sstewart6: that section describes how to deal with this right now
<lukeis> example of extending the protocol: https://github.com/selendroid/selendroid/commit/d275f3a10b63b3fb9616f0bc677b3262addd6c95
sstewart6: I will attempt add a [placeholder] for logging API [soon]
JohnJansen: in 4.1 about creating
... shoudl that autoatically lauch a browser window?
... or can it connet to an existing window?
sstewart6: turns out that connecting to an existing window only can be made to work well in IE
[mention that it's possible with Marionette]
mdas: you mean browser window or tab?
sstewart6: new browser
... new session brings up an independent session
mdas: in this case we have someone who wants to connect to a particular tab
sstewart6: so they need to know
the session id
... just generate one
... there is not an existing command but there are ways to do it
... the thing is of course you need to start the browser ahead of time in some session that would allow it
[Simon reads from section of spec about generation of session IDs]
lukeis: do we not have an invalid-session error?
sstewart6: please raise a bug for that
JohnJansen: 6.5 Windos
... we need a minimize Window command
... and where does the center fo the window
... would you like for us to define it?
JohnJansen: so yes we would like to be able to test a minimized and then back to maximized
<scribe> ACTION: John to send some spec language for window minimize [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action18]
JohnJansen: what is windowID?
sstewart6: unique handle used to refer to the window
jimevans: on Windows it assigns a GUID
JohnJansen: execute async
... we're not clear that it'S trying to do that execute script can't
sstewart6: avoids the need to do
... just have a callback to get executed
MarcFisher: allows the JS event quite tcontinue
lukeis: we do this in the Selenium source
<scribe> ACTION: Luke to port Seleniium async script test to spec testsuite [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action19]
MarcFisher: there's also somethingn used for testing Angular
JohnJansen: how does one click on something other than the center of an element?
... takes a webelement and two numbers
JohnJansen: modal dialogs
... looks like a lot of stuff has been added
... about not needing focus
sstewart6: depends on the
... strawman is you give the credentials at the start of session
... I've not been able to get it work cleanly when I was experimenting
... but we've gone 7 years without it, seems like we can move it to Level 2
... does that still point to something nonexistent?
sstewart6: yeah you're
<scribe> ACTION: Add definition for getElementText to the appropriate section of spec [recorded in http://www.w3.org/2014/02/25-testing-minutes.html#action20]
JohnJansen: can WebDriver find noscript elements?
sstewart6: yeah if it's in the DOM
AutomatedTester: something we need to talk about,given that Touch Events is being replaced by Pointer Events
<sstewart6> I've added some agenda items for tomorrow: https://www.w3.org/wiki/WebDriver/2014-February-F2F
mdas: don't know about the order of events so it's not something that should be in teh spec
sstewart6: browser knows the order of events
AutomatedTester: activation triggers based on DOM Events spec
JohnJansen: no gesture?
sstewart6: multi-action supports it
AutomatedTester: there are differences in what a flick is across browsers
<sstewart6> We should kill "action_chain" in the notes
JohnJansen: can't write a gesture test that works in all browsers?
AutomatedTester: the EndPoint isn't there
JohnJansen: so the right thing to do is the atomic action?
sstewart6: anybody else have queries?
<mdas> lukeis: here's our TouchActions object (for single finger) https://bugzilla.mozilla.org/show_bug.cgi?id=976807
dannin: speed of execution
AutomatedTester: we should be doing things as fast as possible, and the client code [will need to manage timing]
<sstewart6> Those are docs links
This is scribe.perl Revision: 1.138 of Date: 2013-04-25 13:59:11 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/advertiser/advertisers/ Succeeded: s/across browsers?/across browsers/ Succeeded: s/Jo/JohnJansen:/ Succeeded: s/speed up execution/speed of execution/ Found Scribe: wilhelm Inferring ScribeNick: wilhelm Found Scribe: MikeSmith Inferring ScribeNick: MikeSmith Scribes: wilhelm, MikeSmith ScribeNicks: wilhelm, MikeSmith Present: MikeSmith JohnJansen wilhelm seva MarcFisher LukeInmanSemerau ato jimevans mdas SimonStewart samuong DavidBurns ShuotaoGao Dale Annin dannin Agenda: https://www.w3.org/wiki/WebDriver/2014-February-F2F Got date from IRC log name: 25 Feb 2014 Guessing minutes URL: http://www.w3.org/2014/02/25-testing-minutes.html People with action items: add ato ato__ automate_ automatedtester david definition for getelementtext john luke simon sstewart6[End of scribe.perl diagnostic output]