[whatwg] Spellchecking mark III

[Peter Kasting]

2009/1/26 Křištof Želechovski <kkz at mimuw.edu.pl>

>  Q: Should the localization influence the spell checking mechanism?
>
> A: Definitely, since the user is likely to write most messages in his
> preferred UI language.
>
Which is why this is a perfectly valid input for the heuristic the UA uses
to determine the checking language.

> Q: Is GMail a use case for having spell check without specifying a language
> to check against?
>
> A: No, it is not.
>
You don't provide any reason why not.  "The user is likely to write most
messages in his preferred UI language" (which is not true of all users, but
leaving that aside) does not imply "the user will write all messages
exclusively in his preferred UI language".  Therefore gmail cannot
(correctly) specify the spellchecking language of editable fields.
 Therefore the UA must decide.  Unless the probable input language of a
particular field differs from that of the rest of the page, there's no
reason for gmail to specify the probable input language of that field.
 There is no benefit to conflating this concept with "should this field be
spellchecked".

> Q: In case when the user decides to use another language, is the user agent
> free to detect it?
>
> A: Yes, it is, unless the language specified is private, which means the
> field was not intended for checking.
>
Again, this is needless conflation.  You gain nothing, and lose both clarity
and flexibility, by mapping "don't spellcheck" to "specify the language as
private" in this way.  In terms of the semantics of the page, this is
extremely confusing, sicne whether a field should be spellchecked and what
language it's in are nearly orthogonal concepts.

> Q: When the language recognition technology advances to an acceptable
> state, will it be possible to extend the language attribute to explicitly
> request automatic identification of the language?
>
> A: Yes, it is.  Just specify lang="auto" or whatever is agreed upon.
>
There is no benefit to forcing authors to say lang="auto".  What have you
gained?  What if they _don't_ say this?  (The HTML5 spec must still say what
the UA behavior is.)

Language detection libraries today are already extremely good, far more
reliable than anything explicitly set ahead of time by authors _or_ users.
 Unless I am completely misunderstanding you, I think your suggestions fail
to solve the original use cases for the spellcheck attribute, add needless
burden on web authors, and would be completely ignored by UAs who wished to
provide a good user experience.

PK
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.whatwg.org/pipermail/whatwg-whatwg.org/attachments/20090126/5dfb771a/attachment.htm>