HTML validators - are they reliable?

[Andy Dingley]

Call it a primitive tool if you want. It doesn't matter. A hammer is still a
"primitive" tool but still quite useful.

You can drive screws in by hitting them, but that doesn't make a hammer
into a screwdriver.

You're selling a "HTML lint" and calling a validator. "Validator" has
a meaning, and yours isn't it. Yes, you can use your tool to improve
web sites, and even to help them become valid. It doesn't do it by
validating them.

Have you notified the W3C that there's no such thing as an "HTML Validator"?

I presume Jukka's point is that validation is fundamentally defined by
the SGML protocol (or XML) and isn't a specific feature of HTML.
Validators are thus "SGML validators" not "HTML validators". Of course
there can be a HTML validator - it's just an SGML validator that only
works for the HTML doctypes. It's a subset of the same SGML
functionality though, not anything new or specific to HTML.

A HTML validator that is _not_ an SGML validator is not a HTML
validator, it's only an approximation. It might even only be a partial
SGML validator (plenty of SGML features are not required for HTML) and
it could be a HTML validator, but it still needs to follow the basic
algorithms of SGML validation and to base its validation "process" on
the rules expressed by the DTD, not by the developer's own ideas of
"what's important".

 

[Jukka K. Korpela]

Scripsit Albert Wiersch:

Have you notified the W3C that there's no such thing as an "HTML
Validator"?

Of course.

It seems they have a link on their front page called
"HTML Validator" and have had that for quite some time.

They also say (on the page pointed to by the link):

" This is the W3C Markup Validation Service, a free service that checks
Web documents in formats like HTML and XHTML for conformance to W3C
Recommendations and other standards."

The first part is more correct, and then they go again. But at least they
are not cheating people to pay to them for something that isn't what they
claim it to be.

It seems like you live in your own
world with your own definitions.

I didn't invent the concept of markup validator. I didn't invent the concept
of honesty, but I still try to be honest, unlike so many others.

Again, you continue to say more things that aren't true.

That's what you say, but you have never managed to refute the statement I
made.

And you intentionally mislead people in order to make more money by selling
your product.

 

[Albert Wiersch]

You're selling a "HTML lint" and calling a validator. "Validator" has
a meaning, and yours isn't it. Yes, you can use your tool to improve
web sites, and even to help them become valid. It doesn't do it by
validating them.

Actually, if you want to be so technical, then the new version of CSE HTML
Validator (v8.0) does include a validator per the definition you use. Of
course there are other definitions and meanings of validator and, using the
commonly accepted definition (though not accepted by all), CSE HTML
Validator has always been a validator.

Albert

 

[Albert Wiersch]

That's what you say, but you have never managed to refute the statement I
made.

That's because I realize that some things, no matter how true, some people
will never believe and it's not worth refuting or arguing about. But the
truth is still the truth whether or not people believe it.

And you intentionally mislead people in order to make more money by
selling your product.

This is an example of one of those things that is untrue that I can't
convince you that it is untrue. I understand that there is nothing that will
change your view on it. Though perhaps knowing that v8.0 std/pro now
includes a "real validator" according to the only definition you understand
will help?

Albert

 

[Harlan Messinger]

As for the original question, HTML validators are great tools that should
always be used, but they're not perfect. They can't tell you that a page you
wrote will display perfectly as intended.

That's like saying toothbrushes are great tools that should always be
used but they're not perfect because they don't shave you or clip your
nose hair.

 

[Albert Wiersch]

That's like saying toothbrushes are great tools that should always be used
but they're not perfect because they don't shave you or clip your nose
hair.

I don't see how that makes sense. I think a better analogy would be:
"That's like saying toothbrushes are great tools that should always be used
but they're not perfect because they can't guarantee that all plaque was
removed and that your mouth is perfectly clean."

Should you still brush? Yes. It helps.

Albert

 

[Harlan Messinger]

I don't see how that makes sense. I think a better analogy would be:
"That's like saying toothbrushes are great tools that should always be used
but they're not perfect because they can't guarantee that all plaque was
removed and that your mouth is perfectly clean."

No, because while plaque removal is one of the goals of brushing one's
teeth, the features whose absence you consider to be imperfections in
the validators have nothing to do with validation. "Validation", no
matter how many times you say otherwise, has a specific definition in
this context because it's defined in the specification by virtue of
which this context even exists. The information provided by your tool,
no matter how useful it may be, falls outside of the purview of validation.

 

[Albert Wiersch]

No, because while plaque removal is one of the goals of brushing one's
teeth, the features whose absence you consider to be imperfections in
the validators have nothing to do with validation. "Validation", no
matter how many times you say otherwise, has a specific definition in
this context because it's defined in the specification by virtue of
which this context even exists. The information provided by your tool,
no matter how useful it may be, falls outside of the purview of
validation.

Well, it depends on your definition of validation and why people validate...
MOST people validate because they are writing HTML and XHTML to be viewed
with browsers by visitors who come to their site. One of the purposes of
validating is to help make sure their documents do not contain issues that
would affect their visitors and negatively affect how their page could be
displayed. They want their site to display as intended. So how can you say
that validation is not intended to make sure that a document has some level
of correctness which carries over to actually displaying the document in a
browser?

If you are only creating documents so that they "validate" according to an
SGML validator (or dare I say HTML Validator) and you could care less about
how the pages actually display, then I suppose what you say makes sense. But
most people write and validate pages to be seen by real-world browsers. They
don't validate only to pass a strict "test". There's a purpose to validation
which carries over to displaying pages correctly in browsers.

Albert

 

[Harlan Messinger]

Well, it depends on your definition of validation

And again, there is *an* applicable definition of validation, not "your"
definition versus "my" definition.

and why people validate...

Validating is what it is. Either people want to or they don't. If they
want to do something else as well, then they can do that, but then
they're doing that *instead of* validating.

 

[Chaddy2222]

Well, it depends on your definition of validation and why people validate...
MOST people validate because they are writing HTML and XHTML to be viewed
with browsers by visitors who come to their site. One of the purposes of
validating is to help make sure their documents do not contain issues that
would affect their visitors and negatively affect how their page could be
displayed. They want their site to display as intended. So how can you say
that validation is not intended to make sure that a document has some level
of correctness which carries over to actually displaying the document in a
browser?

The main perpus of validating your HTML code useing the proper SGML or
XML passer is to make sure there are no actual errors in the code.

If you are only creating documents so that they "validate" according to an
SGML validator (or dare I say HTML Validator) and you could care less about
how the pages actually display, then I suppose what you say makes sense. But
most people write and validate pages to be seen by real-world browsers. They
don't validate only to pass a strict "test". There's a purpose to validation
which carries over to displaying pages correctly in browsers.

It's not actaully the job of HTML to determan how a page displayes on a
browser, that's what CSS is for, so yes going on what you are saying,
your HTML needs to be properly validated to make sure that the CSS can
then work properly (as intended) and to make sure the CSS works
correctly you validate that as well.

 

[Toby Inkster]

There is no single tool that does everything.

Though Emacs tries.

 

[ato_zee]

Though Emacs tries.

And even if it validates, both css and html, it may be
munged by one or more browsers and may not display
as the creator intended.

I'm wrestling with a similar problem, code validates,
display items mis-aligned on local machine (Firefox and
IE) upload to server, looks fine when opened as a
URL. As both the HTML and CSS are text files, and
view code shows them to be the same as the uploaded
local copies, why the difference?
Same PC, same UA's, view differently.

 

[Toby Inkster]

I'm wrestling with a similar problem, code validates,
display items mis-aligned on local machine (Firefox and
IE) upload to server, looks fine when opened as a
URL.

Something to do with HTTP headers perhaps. (Granted, it's unlikely.)