Re: New issue: error recovery practices (Re: Proposed TAG Finding: Internet Media Type registration, consistency of use)

On Monday, June 3, 2002, 5:37:50 PM, Keith wrote:

>> KM> XML users may be different than HTML users.  HTML is mostly written
>> KM> for eyeballs.  Raise the barrier for XHTML too high, and it won't
>> KM> get used as widely as you'd like.
>> 
>> Counter example - SVG. SVG is also written for eyeballs, has clearly
>> defined error reporting rules that includes of course the XML WF
>> requirement, and i have not heard a *single* complaint along the lines
>> of "can we relax things".

KM> okay, but SVG isn't nearly as widely used as HTML,

Granted, if by HTNKL you mean the stuff web pages are written in. If
you meant the spec, the its a lot more widely used ;-)

KM> nor presumably, as widely used as XHTML would be.

That is a presumption that is difficult to guess at, at this point.

KM> I think there's something about the sheer number of different people 
KM> generating a kind of content that affects the variability of content,
KM> including the liklihood of generating errors.

No, I don't see a difference there. With SVG there are already a lot
of people using a very wide variety of authoring methods from fully
manual to highly automated, or a mixture.

KM> Still, I have no problem with raising the bar - my main point is that
KM> if you're going to complain about errors then you need to arrange for
KM> those complaints to go to somewhere that is useful.

Agreed.

KM>  Having browsers
KM> complain about errors to the user only helps if the developer uses
KM> that particular browser to test the content.

Yes (although the fact that a content developer will use *a* browser
to examine their content is pretty much a given.

KM> Having the browser complain about errors to the origin server
KM> essentially allows the content-provider to do continuous quality
KM> monitoring. Of course, the two approaches aren't mutually
KM> exclusive.

I agree.

Having quality complaints from browsers show up in the server log file
is an interesting concept. There is scope for a whole series of 7xx
response codes, like "not acceptable" ;=)


-- 
 Chris                            mailto:chris@w3.org

Received on Monday, 3 June 2002 14:18:53 UTC