A user account is required in order to edit this wiki, but we've had to disable public user registrations due to spam.

To request an account, ask an autoconfirmed user on Chat (such as one of these permanent autoconfirmed members).

Component Model Discussion: Difference between revisions

From WHATWG Wiki
Jump to navigation Jump to search
(→‎CSS and Rendering: make tab-index and z-order "soon")
(→‎Now: removed layer question)
Line 19: Line 19:
* [http://wiki.whatwg.org/index.php?title=Component_Model_Discussion:_Rendering rendering sub-page]: discussions on rendering details
* [http://wiki.whatwg.org/index.php?title=Component_Model_Discussion:_Rendering rendering sub-page]: discussions on rendering details
* [http://wiki.whatwg.org/wiki/Component_Model_CSS_Brainstorming CSS brainstorming page]: brainstorming on alternative solutions to styling
* [http://wiki.whatwg.org/wiki/Component_Model_CSS_Brainstorming CSS brainstorming page]: brainstorming on alternative solutions to styling
=== <span style="color: #006633">Now</span> ===
* Do components create a layer?
** <span style="color: blue">assumption: no</span>
::[[User:Dglazkov|Dglazkov]]: isn't this an implementation detail?


=== <span style="color: #008866">Soon</span> ===
=== <span style="color: #008866">Soon</span> ===

Revision as of 02:29, 12 October 2011

Overview

This document attempts to exhaustively list issues that we feel are unanswered or ambiguous.

Linked pages:

Issues

  • Now: in the spec draft and/or already in use for internal components
  • Soon: to be included in V1 spec
  • Later: issues to be discussed for later versions of the spec

CSS and Rendering

Soon

  • styling of rendering children of <content> (partially discussed in the CSS brainstorming page).
  • tab-index behavior
  • z-order behavior

Shadow Nodes

Now

  • shadow root being a separate node type vs. an element vs. document fragment vs. ??? (see rendering sub-page)
    • current state: separate node type
  • Can the same host element child be rendered under multiple <content> elements that all select it?
    • assumption: no - the first to select it "gets" it. <content> without select is always last in this regard
  • Is <content> is a bona-fide HTML element (vs. a DOM subtype)?
    • assumption: HTML element

Soon

  • Does ShadowRoot have innerHTML?
    • assumption: yes
  • details of <content> selector syntax
    • suggestion: space-separated list of selector sequences: [tag]?[class]*[id]*[attr]*[pseudo-class]* (NO pseudo-elements)
Dglazkov: will space-separation be confusing? I like the idea, but just thought I should raise this as a concern.
Rolandsteiner: I thought a space-separated list follows HTML conventions more closely, but perhaps to draw an analogue to CSS (and to allow potential later extensions), a comma-separated list may be better (?).


Events

Soon

  • focus event traversing out of shadow?
    • assumption: yes, redirected
    • current state: not implemented

Later

  • instantiation events (which events & when?)
    • for <template> instantiations
    • for component instances
    • on rebinding? (provided CSS 'binding' is possible)
  • event on component registration?

Loading and Life Cycle

Soon

  • loading mechanism (syntax)
  • How to handle definition after use
    • assumption: don't do anything
  • packaging
  • is it possible to rename a component? e.g., to avoid name clashes?
    • assumption: no
  • define all edge cases and “unhappy” cases. Examples: constructor throws exception (general case), constructor throws exception during parsing

Later

  • is it possible to have a scoped (private) names, e.g., for private components?
    • assumption: no
  • Is it possible to undefine a component? (if yes, how?)
    • assumption: no


Registration

Soon

  • how to handle name clashes between components?
    • assumption: teach authors to add unique prefix, e.g., <x-cyberdyne-terminator>
  • details on attaching a shadow tree?
    • <x-mywidget>
      • Pros: flexible, tag is inherently static, most performant
      • Cons: accessibility, fall-back
        • <div becomes=mywidget> even more problematic in these regards and also requires duplication of style rules (?)
    • <button is=mywidget>
      • Pros: accessibility of base element, natural fallback
      • Cons: not a priori obvious that 'is' cannot be rebound, limited to existing element names
    • CSS binding: button { binding: mywidget; }
      • Pros: accessibility and fallback, most flexible
      • Cons: implementation is hard, multiple (perhaps mutually exclusive) bindings may apply, cannot extend the API
    • other?
Dglazkov: Hixie suggested a mix of decorator and element, where the API is bound using magic attribute and shadow tree + events are bound using css binding.
  • fallback mechanisms (implicit with <button is=mywidget>, explicit, e.g., with fallback content as children)
  • Are shadow trees dynamic or static?
    • assumption: static, but see discussion in public-webapps@ ML

Later

  • accessibility (implicit, e.g. with <button is=mywidget>, explicit, e.g., with ARIA roles)


Derivation

Later

  • should we do XBL-like inheritance?
    • assumption: no
  • “aspect-oriented" components (components defining the parts they use, derived components can override or re-use specific parts, part names automatically become pseudos)
  • composition vs. inheritance (i.e., do we need inheritance at all, or can composition handle all use cases?)
    • temporary assumption: composition only

JavaScript

Now

  • Can one subtype anything (text/attribute nodes, events, ...), or just HTMLElement?
    • assumption: just HTML element
  • details of construction

Soon

  • can one write a constructor with <supertype>.call(...), or does one have to use createElement?
  • Isolation/Confinement JS implementation details
  • Forwarding JS implementation details (ex: sub-classing <textarea> (Maciej’s examples), auto-completion, omni-bar like drop-down, toolbar)
  • how to handle Element.register being overwritten?
    • assumption: don't care
  • how to handle, e.g., appendChild being overwritten on a component?
    • assumption: don't care

Later

  • Attribute/property forwarding.

Isolation and Confinement

(see also the isolation brainstorming page)

Soon

  • To what degree is isolation/confinement really necessary? (scripts? DOM access? CSS properties?)
  • How far is isolation/confinement handling for components similar to <iframe>?
  • Can we substitute isolation with <iframe>s?
  • Are instances of the same component isolated from each other?
    • assumption: yes, but they should be able to set up shared data structures
  • Are instances of the components from the same domain isolated from each other?
    • assumption: yes, but they should be able to set up shared data structures
  • Can instances of the same component share data somehow?
    • assumption: yes (how?)
  • Can instances of the same component and its derivatives share data somehow?
    • assumption: yes (how?)
  • Can instances of the components from the same domain share data somehow?
    • assumption: yes (how?)
  • Can isolated components position content outside the host boundary rect? (if not, problems with, e.g., +1 widget)
    • assumption: yes


Declarative syntax

Soon

  • how/when to run scripts?
  • syntax to style the host (provided we select a model where it is rendered)?
  • <template> details:
    • usable without components?
    • instantiable without explicit components (e.g., <instance select=”x-cyberdyne-terminator”>)
    • event handler registration syntax
  • <element> details
  • Do we need <binding> as well?
  • How to define the API of components

Later

  • <template>: introduce template types (e.g., <template type=”django”>)?
    • assumption: no


Misc.

Now

  • How can we split the spec into (more) different parts?

Later

  • Do we define the shadow DOM (or parts thereof) of built-in elements?
    • assumption: no