"...you can’t have a standard with just a spec–you must have a super-strict reference implementation, and everybody has to test against the reference implementation. Otherwise you get 17 different “standards” and you might as well not have one at all."
Ok, there has to be One Authority, I buy that. Joel implies that since we don't have that now we can never have one. Maybe I'm being an idealist in my own way, but why can't we have this super strict implementation? The w3c has some resources, right?
Can't the standards people pressure them to create this reference implementation? It doesn't have to be fast, or have any features or anything. It just needs a viewport, back button, a way to point it at an url and to be Official. (also, scriptability would be awesome). It wouldn't fix old web pages, but it could stop the spread of new problems.
Hmm. On second though, it's probably more likely that the w3c would give us all ponies...
How about this: nobody use html 5 until we've got our reference. Promise?
4 comments:
"Can't the standards people pressure them to create this reference implementation?"
You mean like, the various test suites that have been long excepted? MSFT even submitted a few to CSS2 to make sure their implementation was good in the IE8 beta.
Joel's a bit off base, tbh.
See also: http://diveintomark.org/archives/2008/03/18/translation-from-ms-speak-to-english-of-selected-portions-of-joel-spolskys-martin-headsets
So, those are good things. The html5 parser project has plenty of good people involved. But I still like the idea of an 'Official' validator.
Obviously, the CSS one is official, but the CSS2 test suite page says this (not inspiring confidence (CSS2 has been around a while)):
"This is a development version of the CSS2.1 Conformance Test Suite.
It is woefully incomplete and contains many incorrect tests. Please check the latest version of the CSS2.1 specification and the errata list before assuming a failure is due to an implementation bug and not a test suite bug."
Of course, it's all moot if MS just ignores the whole thing, as diveintomark claims (tbh, I haven't paid enough attention to the working group activities to know he's right, but I have no reason to doubt it.)
do you mrean a validator or a conformance checker? :)
that's part of the problem right there though - everyone wants something different.
What I would like, in an ideal world, is some way to say definitively that a browser was rendering a page 'correctly'. Maybe that is better described as a (browser) conformance checker.
Post a Comment