[whatwg] Spellchecking mark III

Ian Hickson ian at hixie.ch
Fri Feb 13 12:09:56 PST 2009

On Thu, 12 Feb 2009, Kristof Zelechovski wrote:
> Regarding http://html5.org/tools/web-apps-tracker?from=2800&to=2801, my 
> requests:
> 1. Change the literals "true"/"false" to "on"/"off", leaving the DOM values
> Boolean.

There are three of these attributes so far:

  autocomplete = on/off
  contenteditable = true/false
  draggable = true/false

I used true/false for spellcheck since it had slightly more other 
attributes doing the same thing.

Also, it's been implemented twice now, so using other keywords is a 

> 2. Check the spelling of the passage ("asits"!) :0)


> 3. Say that the default behavior for BODY is "on" and the default behavior
> for INPUT[type=text] is "off".

The default behavior is user-agent-dependent. This is intentional since 
different users may have different needs.

> 4. (I understand that it is implicit that this "SHOULD indicate" does 
> not make tiny clients that do not have the resources non-compliant?)


> Stretching it a bit, a user's language always matches the site's, 
> otherwise the user would not be able to submit to the site anything that 
> makes sense, except when the site is a gateway for submissions to an 
> uninvolved third party, in which case said submissions should be tagged 
> with the language of submission anyway (IMHO).

On Thu, 12 Feb 2009, Bil Corry wrote:
> Let me give you an example where this isn't true.  I'm in the United 
> States and I do contract work for a company in Germany.  At the German 
> company, they have an internal bug tracker for their intranet 
> applications.  Usually the bug descriptions are written in German, 
> except mine, which are in English.  So they will submit bugs in both 
> German and in English, depending on who is taking care of the issue.
> How do you envision the UA will determine which language the user is 
> writing in?  And what happens when the user submits both German AND 
> English, for two audiences?

On Thu, 12 Feb 2009, Kristof Zelechovski wrote:
> The server has two ways of knowing the user's preferred language: the 
> user's preferences and the browser settings, in that order.
> Submitting in two languages usually needs two controls, one for English 
> and one for German, with appropriate markup.  The server must be 
> prepared to handle this use case.

On Thu, 12 Feb 2009, Aryeh Gregor wrote:
> Both of which are often wrong.  Users may be multilingual, and multiple 
> users may use the same computer.  On the forum I administer, I post 
> almost exclusively in English.  However, sometimes I find occasion to 
> write a post partly or wholly in Hebrew.  How is the site supposed to 
> know when I'll decide to do that before I even start typing the post?  
> How can the site ever be sure what language the user will type until he 
> actually starts typing?
> The server might be able to make an educated guess as to what language 
> will be entered, but so can the browser.  And the browser is in a *much* 
> better position to check that guess, because it has access in real time 
> to the actual text the user is typing, plus the user interface language, 
> and -- of course -- any lang= or xml:lang= attributes specified in the 
> HTML.  Ergo, the logic should be left up to the browser.

On Thu, 12 Feb 2009, Kristof Zelechovski wrote:
> The language attribute can be changed at run time if needed.  It 
> requires an additional event that can be called "langmismatch".  Of 
> course, a more traditional selector is also a solution.  If the site is 
> primary English, with Hebrew fragments here and there, it is not much 
> harm that the fragments are considered spelling errors (although, in the 
> case of English/Hebrew bilingualism, it is unlikely because the 
> character set is different). In short, the user agent is allowed to use 
> whatever AI it is equipped with.
> Markup for German AND English submissions at the same time, as per your 
> request:

On Thu, 12 Feb 2009, Bil Corry wrote:
> In my case, we have a single field, "bug description" that may contain 
> both English and German.  And in some cases, even a pure German bug 
> report may reference the English form fields, such as:
> 	Legen Sie "City" vor "Postal Code"
> In that case, there is no way for a UA or Server to auto-determine the 
> language, even if you're aware the user speaks both German and English.
> My suggestion is to leave the lang attribute out of the spec, and let 
> the UA handle it as it wants.

On Thu, 12 Feb 2009, KÅ~YiÅ¡tof Želechovski wrote:
> Having interjected words marked as spelling errors is not a failure.  
> The same phenomenon occurs with proper names and you cannot help that. 
> The UI you described is inconsistent and it should be fixed.  The 
> control for German should be labeled "Fehlerbeſchreibung" or whatever.

On Thu, 12 Feb 2009, Kristof Zelechovski wrote:
> I do not know much about UI standards but the rule that the answer 
> should be formulated in the language of the question is rather 
> straightforward.  It is just common sense.  Exceptions are questions 
> like "How is that in German?".

On Thu, 12 Feb 2009, Bil Corry wrote:
> No one can control the language a user will choose to use in a textarea, 
> regardless of the label used to describe it.
> Providing a localized textarea for every language might increase the 
> odds of the user using the language the server prefers, but there is no 
> guarantee.  And I'm unclear what problem that would ultimately solve.

On Thu, 12 Feb 2009, Kristof Zelechovski wrote:
> The majority of users will answer the question in the language of the 
> question, this is the normal reaction.  Of course there is no guarantee 
> but the odds of getting the expected result are high.  Assuming that the 
> user's input will actually be read by somebody, providing proper markup 
> will help the readers to get something they are able to read.

Based on the above feedback, I have left the spec as-is, allowing both 
behaviors ("User agents may use the language of the element to determine 
what spelling and grammar rules to use, or may use the user's preferred 
language settings").

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

More information about the whatwg mailing list