- From: Richard Schwerdtfeger <schwer@us.ibm.com>
- Date: Mon, 31 May 2010 09:33:44 -0500
- To: HTML Accessibility Task Force <public-html-a11y@w3.org>, cyns@exchange.microsoft.com, faulkner.steve@gmail.com
- Cc: david.bolter@gmail.com
- Message-ID: <OF7E8B3344.CBB9D69A-ON8625772E.004200B8-86257734.004FFE5A@us.ibm.com>
Below is my assessment. Before I do any further accessibility API mapping (this will need to be a separate document from the HTML 5 spec.) we should go through and discuss my assessment on Tuesday. There are MANY Dom methods provided for each of these widgets that provide information beyond what is provided in the DOM attributes. Do we need to do anything with that information when dealing with the HTML 5 accessibility API mappings? That would require HTML 5 to allow ant AT or the user agent to call these methods. That is a new wrinkle and has the group considered that possibility. Also, I have identified some issues with the HTML 5 specification we should discuss. For all of these roles the author should NOT be able to override element with an ARIA role. 1. details element Implied Native semantics: role="button" aria-labelledby="id" of the summary element open content attribute = aria-expanded DOM children of details: <summary> provides the level of semantics as a div. example: http://www.whatwg.org/specs/web-apps/current-work/multipage/interactive-elements.html This may be overriden by a combobox as there is no reason the author could not convert this to a button with a drop down menu. Here anything in the details area could be selected as a menu item. This would change the functionality of the control. 2. meter is an range object in the ARIA taxonomy Implied Native semantics: role="progressbar" It is defined as a scalar guage and not a control meaning it should not become a slider or a scrollbar aria-valuemin = min content attribute aria-valuemax = max content attribute aria-valuetext= the content within meter is the equivalent low, high, and optimum are not ARIA values and we need to make sure they are exposed through the accessibility API on each platform. Also, some notification that properties exceed or fall below a particular level should be created. This should not be overriden with ARIA properties as it will create conflicts 3. The user agent can bind whatever they want to this input type. So all though there is a constraint on the data type implied semantics in terms of UI components are indeterminant. input type=date – date input control input type=month – year-and-month input control input type=week – year-and-week input control input type=number – number input control input type=color – color-well control Implied ARIA semantics: These accessibility mapping is dependent on the browser implementation. So, these input types have NO determinable ARIA mapping. The ARIA mapping and accessibility API mapping are dependent on the control bound to the input element used by the browser. These do all support a min and max value. However, the min and max value could be strings vs. numerical values. One can also imagine that the user agent will draw a different control based on the real estate available to the device. 4. input type=url Implied Native semantics: Although the rendering is not stated, this is a textbox. The ARIA role for this is textbox. 5. input type=email – e-mail address input control Implied Native semantics: Although the rendering is not stated, this is a textbox. The ARIA role for this is textbox. 6. input type=range – imprecise number-input control Implied Native semantics: The ARIA role for this is a slider. The aria-min, aria-max, and aria-valuenow values should be equivalent to the min, max, and value content attributes provided in HTML. Aria-valuetext should be allowed on this element. There is a DOM method corresponding to returning a string equivalent for all these content attributes. 7. input type=search field Implied Native semantics: The definition of what is to be rendered is unclear. No ARIA role can be determined. 8. input type=time – time input control Implied Native semantics: The ARIA role for this is a spinbutton. The aria-min, aria-max, and aria-valuenow values should be equivalent to the min, max, and value content attributes provided in HTML. Aria-valuetext should be allowed on this element. There is a DOM method corresponding to returning a string equivalent for all these content attributes. 9. input type=datetime – global date-and-time input control and input type=datetime-local – local date-and-time input control Implied native semantics: ARIA has no direct role for the combined control. This is a combination of a date and time control. Now the Opera rendering of this separates date and time into two controls, yet the values are integrated (max, min, value) into single values. Should we ask that they be separated? Within the user agent this could be mapped to a group role with child date and time (spinbutton) objects. The values for max, min, and current value should be split between the two controls. The actually ARIA role for date is non-deterministic based on the spec. The time control is. Internally, the browser should create a role=group. where the label could be datetime but actually should be supplied by the author. There should be a defect logged requiring a label for the date and time control and date and time local control. 10. progress – progress indicator. Implied native semantics: The role is "proggressbar" aria-currentvalue is position aria-valuemax is max aria-valuemin is 0? Authors should be able to supply aria-valuetext There should be a defect log to have a minium value or to at least state in the specification that it should be zero or undeterminable. There should be an issue raised regarding what resides between the start and and end <progress> elements. Is it fallback content? This is unclear. 11. input type=tel – telephone-number-input field Implied native semantics: The role is "textbox" References Used: http://www.quirksmode.org/html5/compatibility.html http://dev.w3.org/html5/spec/rendering.html#rendering http://dev .w3.org/ht ml5/spec/rendering.html#bindings http://dev.w3.org/html5/markup/ This web site shows that even the Opera people do not know what to do with many of these and that man of the browser manufactures do not have inplementations: http://www.quirksmode.org/html5/compatibility.html Rich Schwerdtfeger CTO Accessibility Software Group
Received on Monday, 31 May 2010 14:34:40 UTC