Thursday, 6 December 2012

Change to HTML Specification

I made a comment concerning whether lists are ordered or not. As a results Ian Hickson made a change to the HTML Specification in the section Splitting a String on Spaces.

This is the commit.

My name is now in the acknowledgements.

This is the original post.

Hello. I'm part of the Seneca College student who are implementing the
WebVTT spec for Mozilla.

The syntax rules state, "Each component must not be included more than once
per WebVTT cue settings string." The parsing rules allow for duplicate
settings implicitly because it makes no such check. The splitting on a
space creates a list, not an ordered list. That means a cue with duplicate
settings will parse but with unpredictable results.

Question. Should this simply be left as unpredictable behaviour? Or should
the parser make the behaviour predictable? If the settings could be
tokenized in to an ordered list then either it could use the first or last
duplicate setting (last should be easier).

My feeling is that it should remain unpredictable because the syntax states
only one is allowed and therefore duplicate settings should not be
supported. If this is the case, it might be useful to have a note in the
setting paring section stating that while duplicate settings will be
parsed, the behaviour will be unpredictable.

Thank you,
Kyle Barnhart

My second post.

Let me point out why I asked this question.

Twice in the WebVTT specification it is explicitly pointed out that an
"ordered list" is used.
-  WebVTT Internal Node Objects have an ordered list of child WebVTT Node
Objects [1]
-  WebVTT Internal Node Objects also have an ordered list of class names [1]

In addition the word "append" is used 27 times in the specification
indicating that order is important.

In the parse WebVTT settings algorithm [2], it only states that there is a
list of setting token. There is no reference to order.

In the split a string on spaces algorithm [3], it specifies only that a
list is created. It also states "add" to the list and not "append". The
common micro syntax document [3] uses the word "append" 9 times. In
addition there are explicit instruction for ordered and unordered sets of

The order of a list is so often explicitly stated or implied through the
phrases "ordered list" and "append" it appears to be the precedent that
lists which must be ordered must use those phrases. I looked very hard for
some sort of clarification that lists are ordered by default but I could
not find any.

If explicitly stating or implying that something is ordered is the
precedent, then when it is not explicitly stated or implied, I must assume
that it may not be ordered.

What it means is that I could create the list from adding to the start or
inserting to the middle or some other method. Or use some complex data
structure that adds key value pairs by some means that does not preserver
order. All of this would be perfectly valid according to the
specifications, but would mean that the order of the
list across implementations is unpredictable.


Ian Hickson's response.

I've made it clearer that these lists are ordered, just so that there's 
no ambiguity here.


See for the most up to 
date copy of this text.

(In particular, WebVTT references the terms as defined in the WHATWG HTML 
spec; I make no guarantees that the references make sense if you use the 
W3C HTML5 spec for the HTML terms.)

Ian Hickson               U+1047E                )\._.,--....,'``.    fL       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

No comments:

Post a Comment