Re: Thanks a lot

Kynn Bartlett wrote:
> [...lots of sensible stuff deleted ...]
> HOWEVER, the message you get when you submit a page that was valid
> last month doesn't say, "hi, we had a buggy validator, which gave false
> positives to millions of web pages, and it's fixed now" -- it says "hi,
> this author doesn't know what the hell he's doing, and is producing
> invalid code."
> 
> [...]
>
> What should be done?  
>
> [...]
> 
> >It also begs the question: "If I fix this, how do I know the same thing
> >won't happen again, with reference to some other aspect of my sites?"
> 
> Exactly. This is why the correct behavior should be to find a way
> around the problem, for example, changing the validation error
> message to a conditional validation or something, instead of simply
> stating that the page does not validate.

Another possibility is that older validators remain online for a given
time.  Consider this hypothetical example:

Hypothetical: Current version is 1.1.6, 
  last version was 1.1.3, 
  version 1.0 was released over a year ago

1) Validators' URLs are version stamped:
  v1.1.6 is found at /check/1.1.6/referrer/,
  v1.0 is found at /chec/1.0/referrer/,
  /check/referrer/ gives a redirect to the latest:
/check/1.1.6/referrer/

2) Validators are kept around for some arbitrary time, ie a year, after
they've been superceded.

3) The "HTML OK" icons and hrefs given by a validator are
version-specific, so the 1.1.6 validator, despite being the newest and
most recent, will give /check/1.1.6/referrer/

4) outdated validators continue to validate the asme pages with their
version (for example, 1.1.3) but give some notice: "This validator has
been found to be inaccurate, but this page is still valid by this
parser.  This validator will be unavailable after 2002-07-14"

5) outdated (greater than the arbitrary cut-off point) give the message
"Validators over a year old (ie less than 1.1.3) are no longer
considered authoritative; please re-validate with validator.w3.org" 
Note: no "this page is bad", just "please revalidate".


OK, obviously a bit of framework, but it would make it so that:
1) if your page said "valid", it'll says "valid" for at least a year (or
other arbitrary time) from any warning or notice of change in the
validator.
2) you'll know in advance when a validator version will be unavailable
3) you'll have a year (or some other arbitrary length of time) to fix
your pages.

Basically no surprises, time to react, no lost "face" anywhere.


There's my $0.02, served with the grain of salt appropriate for an
opinion from a guy who *wouldn't* have to implement it.

Allan

Received on Monday, 10 December 2001 17:47:50 UTC