Here's a chunk of a longer piece of punditry I'm working on, re: the
choices between doctypes for authoring and the State of the Union for
validity. I've got a bucketload of nasty code and a bunch of developers
who need education on validation, validity and all the usual good
stuff. In particular it's a problem of triage on the legacy code:
what's correctable now, what's going to need serious work, what's going
to be a total re-write.
I'd appreciate any comments you might have
Levels of Validity
Formally there are no "levels of validity". Validity is objective and a
document is either valid or not. For authoring we should keep to this
view. For assessing legacy documents though, it is useful to invent
some "levels" of validity. Remember that these are an arbitrary
invention with no formal definition and no standard comparability with
other projects or W3C recommendations. Validity can only be defined in
terms of a specific doctype, so these are defined against Strict.
Trivial invalidity
Use of the <a target="..." attribute. These would be valid
Transitional documents.
Documents are known to render exactly like valid Strict documents .
Minor invalidity
Inline elements placed directly into <bodyor <formelements. Use of
<fonttags. These would be valid Transitional documents.
Documents are known to render like valid Strict documents, because
browsers(sic) compensate(sic) automatically(sic) for these
"well-known"(sic) invalidities.
Major invalidity
Documents that are not even valid as Transitional. Typically this is
caused by improper nesting or closure of elements.
Rendering of these documents relies on SGML parsing producing a correct
DOM of the document elements. This is not a reliable authoring process:
although the SGML behaviour is deterministic, the effects are not
human-friendly and may give quite different elements from the tags that
the author used (particularly for invalid nestings).
It's even possible that "valid" legacy documents should be treated as
being majorly invalid. Where authors have mis-used the element closure
rules, the SGML behaviour is still clear for how this translates to the
DOM elements, but it may not be what the author expected. It's
difficult to identify these valid but incorrect documents, or to tell
some majorly and grossly invalid documents apart, without having
knowledge of what the author originally intended. This is why
automatic-correction of these afterwards is always a hard problem [12].
Grossly invalid
Tag-soup documents that barely resemble the DTD. These are still the
most common legacy documents.
Element set
There are massive advantages to abandoning "HTML 3.2" style layout with
its plethora of <fonttags, <tableused for layout and even the
single-pixel gif. These advantages come about by using the right
elements and ignoring the bogus ones though, not by any magic choice of
doctype.
Although it's true that using a Strict doctype will force you to do
this, it's just as easy to do so with a Transitional doctype. You can
write just as good markup under Transitional as you can under Strict,
it doesn't _require_ you to use <fontor to misuse <blockquotefor
indents.
Don't be afraid of Transitional. It's just as "good" as Strict, so
long as you aren't actually using any of the undesirable elements.
Equally, if you aren't using them then you could use the Strict
doctype. However, if the page is valid as Transitional but not as
Strict, then treating it as Transitional would allow the simple
objective validity test to be passed.
Transitional Differences
Transitional represents the shift in design philosophy from HTML 3.2 to
HTML 4.0 and CSS. The concept behind the two doctypes is that older
HTML 3.2 documents may simply be re labelled as HTML 4.01 Transitional
and that all new documents will be authored as HTML 4.01 Strict.
Transitional was only ever a temporary transitional phase and that was
7 years ago!
Legacy documents
We must accept that we have legacy documents, and that they'll be in
less than ideal form.
Legacy documents should be treated as invalid HTML 4.01 Strict
documents and should have an appropriate doctype declaration added
immediately. Then work on refactoring their validity, where possible.
It's better to do this than to leave them under the old doctypes: even
having valid HTML 3.2 wouldn't be a valuable distinction.
There will be many problems with legacy documents. They're likely to be
invalid, to use a poor set of elements and even to be not well-formed.
The difficulty of bringing them up to a "best practice" standard
depends on how bad particular examples are. Practically, well
formedness should be rapidly achievable, element use and validity may
be impractical to change.
In the rare case of almost-correct legacy documents (trivial or minor
invalidity, as described above), there could be a benefit to using a
Transitional doctype (see the discussion for new work). Otherwise it's
always best to label them as 4.01 Strict and at least gain the CSS
rendering benefits from it, rather than to label them as something
obscure albeit valid under that description. The web (as a consumer of
HTML, rather than an authoring problem) cares less about doctypes than
it does about content.
