Re: Balancing the myth-busting.

On Tue, Aug 09, 2005 at 08:55:24PM +0100, Gez Lemon wrote:
> > You already have done so this evening.
> 
> I've already put words in your mouth, or I've already said that
> education isn't important? 

The former.

> I wasn't aware I had done either. Could you please be more explicit
> so that I can either apologise and retract my statement, or defend
> what I actually said.

http://lists.w3.org/Archives/Public/w3c-wai-gl/2005JulSep/0304.html

"So we agree there's a problem. The only difference is that I would
like to see the problem addressed."

> I explicitly agree with your point of view regarding education. I just
> disagree that there isn't room for improvement with the markup
> validator.

I decided to take a look at the script you pointed to earlier as an
example of why you think the feature is needed. There are lots of
things in it which would be hard to implement to implement in a
validator.

1. Uses external resources. Possibly over http only.

2. Tests user agent string. OK, so it doesn't actually output it into
   the document, but you would have to test every single branch of
   every if statement. This would probably end up with several dozen
   documents to test (and the "JavaScript is disabled" document).

   That would be very expensive in terms of resources, and lead to the
   problem of how that information gets presented to the user.

3. Uses innerHTML. Is the validator supposed to understand
   non-standard DOM methods? Or is it supposed to complain about them?

4. Looks at navigator.plugins and window.ActiveXObject. Is it supposed
   to support these too? 

5. Gets the version of flash installed, assembled from several bits of
   data, then tests if it is high enough. The validator would have to
   support Flash for this. In fact it would have to support lots of
   different versions. Either that or ignore the whole thing and do
   the branching again. This time it uses Trinary syntax though.

I suppose it is just about possible to do what you want. It would have
three major problems though.

1. It would be very difficult and require a /lot/ of man hours of work
   to create

2. It would be very difficult to present the information to the user
   since there are dozens of different documents outputted by the
   script that would be validated.

3. It would be very expensive (in CPU terms) to run

The benefits do not even come close to outweighing the costs in my
eyes.

-- 
David Dorward                                      http://dorward.me.uk

Received on Tuesday, 9 August 2005 20:12:09 UTC