RE: [techs] Test 49 Suggestion

> From: Tex Texin [mailto:tex@i18nguy.com] 
> Sent: 14 February 2005 18:45


Just a few quick notes, since I think other questions you asked should have
been addressed by the document I suggested you read.

> 2) The larger web servers are concerned with the size of 
> their web pages and want to minimize them to maximize 
> performance/throughput. Seems very wasteful to have a 
> nullifying statement added to a document where no statement 
> would do. Since people follow the trends they see in the 
> pages of the larger web servers (meaning those that serve the 
> most pages, like search engines) it would be good if we 
> minded size and made it likely they would comply.

The hope is that if it is required (in XHTML2) authors will find out about
it and use it, rather than just ignoring it, as they can all too easily do
now. If we don't require the attribute, people won't use it and we may never
be able to bootstrap useful language applications. 

I don't think we should worry about one small attribute & value per
document, when we look at the potential benefits to be gained.  

The cases where there is a null value are likely to be extremely rare.  The
rest of the time the attribute will be performing a valuable function.

> 4) If you require an attribute on the entire document, 
> doesn't it override the http setting of language? (not that 
> it works all that well to begin with...) 

That's what we're trying to fix.

> 5) For some documents, where the contents are dynamically 
> created, the language isn't known until later, and so are 
> designed to have a neutral header and then processing fills 
> in. This requires a change for the body generation to also 
> generate the html element now, or to lie about the language 
> (using null or some default value) and assert a new one when 
> the actual generation occurs.
> 
> So although the guideline applies to many documents, it is 
> wasteful and can require rearchitecting the design of 
> dynamically generated documents. I wouldn't make it a requirement...

If the language is not known, then the null value is accurate, not a lie.


RI

Received on Monday, 14 February 2005 19:42:35 UTC