RE: Identifying touch events that can trigger scrolling

At first glance defaultStartsManipulation seems reasonable (not crazy about the name, but even less crazy about bikeshedding over it). I want to take a look at some of the pages we’ve seen using crude JS gesture recognizers & magic numbers to determine if a manipulation is going to occur and see how easy something like this is to swap in.

-Jacob

From: rbyers@google.com [mailto:rbyers@google.com] On Behalf Of Rick Byers
Sent: Monday, January 19, 2015 8:32 AM
To: public-touchevents@w3.org
Cc: Mustaq Ahmed; Matt Gaunt; Jake Archibald; Daniel Freedman
Subject: Identifying touch events that can trigger scrolling

In last week's TECG call<http://www.w3.org/2015/01/13-touchevents-minutes.html>, we agreed that the below problem of identifying touch events that can trigger a click is interesting mostly for browsers that don't have an API for disabling the click delay and that adding such an API is probably a superior solution for those scenarios.

However, the very related (often inverse) problem of identifying touch events which can cause a scroll to start is still very interesting for all browsers.  We liked Ben's proposal<https://bugs.webkit.org/show_bug.cgi?id=138151> of adding an API to touch events for indicating what actions they can trigger, but prefer to start small focused on concrete actions they can trigger.  In particular, an API defined in terms of touch gestures (like "pinch" and "double tap") is unlikely to be standardizable in the current IP climate.

Based on this discussion, I've written up a concrete proposal<https://docs.google.com/a/chromium.org/document/d/1Rf-WafO91JehdjNVGRAkOQHCSYbFju8M90A4Xt1s-A8/edit#heading=h.ay2pmk927mpu> for addressing the "when will scrolling start" problem.  Essentially this adds a boolean to touchmove events called "defaultStartsManipulation" (with some connection to the touch-action specification).

Jake/Matt: in particular I expect this to address the concern<https://groups.google.com/a/chromium.org/forum/#!searchin/input-dev/mattgaunt/input-dev/Bo5FyTLljwY/aZVWhHV04AwJ> you raised awhile back about custom side-swipe and the lack of touchcancel events - especially if we can get a common API supported in all major browsers.

Feedback (here or as comments in the doc)?
   Rick

On Mon, Jan 5, 2015 at 12:48 PM, Rick Byers <rbyers@chromium.org<mailto:rbyers@chromium.org>> wrote:
Sometimes it's useful to detect a 'tap' gesture from JavaScript from the touch events, before the mouse events (click, etc.) are fired.  For example, popular libraries like FastClick<https://github.com/ftlabs/fastclick> need to do this for browsers that don't have a better way to disable the 300ms click delay (but there are other scenarios as well).

Many apps / libraries implement a simple heuristic for this like "look for a first-finger touchstart, followed by 0 or more touchmoves which are less than N pixels from the start, then a touchend" This is relatively simple and works alright in practice most of the time.  But the developer's intention is that it should match the browser's tap detection logic EXACTLY.  If the logic mis-identifies a tap, then you could get double handling (JS behaving as a tap but the browser triggering a scroll).  If, on the other hand, the logic fails to identify a tap it may feel harder to activate a control, or activation behavior may be inconsistent.

There are a number of reasons why such application logic can never be perfect:

  1.  The value of N can vary.  Eg. on Android, device OEMs can control the ViewConfiguration::getScaledTouchSlop<http://developer.android.com/reference/android/view/ViewConfiguration.html#getScaledTouchSlop()> value used by browsers to implement tap detection.
  2.  The browser may use an algorithm more complex than a simple bounding box.  Extreme example: in iOS UIWebView (and WKWebView) the native application has control over the gesture recognizer graph and can arbitrarily influence detection logic.
All in all, it seems web developers should really have a way of asking "will this touchend cause a click event to be generated if it's not canceled?".

I discussed this problem a bit with Benjamin Poulin on the Safari team here: https://bugs.webkit.org/show_bug.cgi?id=138151.  He proposes a property on TouchEvent ("eventCanStartDefault") which somehow indicates which gestures would be started by the default action of the event ("Tap", "Pan", "Pinch", etc.).  I'd be happy to add something like this to chromium as well (due to a recent re-architecture it's actually quite easy for us to implement).

Thoughts?
   Rick