Posted by Andy Dingley on 01/29/08 16:50

On 29 Jan, 15:38, Jeff <jeff@spam_me_not.com> wrote:

> > _Inside_ a CMS, there are strong arguments for using XHTML (or at
> > least, some XML that shares the same XML schema)
>
> > On publishing from a CMS, it's usually easier to serve the document to
> > the web as HTML than it is as XHTML
>
> That has been my thinking. To create it as xhtml and serve it as html.
> That leaves me with either fixing the stray bits like <br /> or just
> ignoring them as the browsers do anyway. I suppose I should fix them...

In general, you just shouldn't even think about writing output
routines!
Really. Stop it right now.

Why do you need to write a "serialiser" at all? You're probably
working with XML, from a well-known language. In which case, there's a
range of XML DOMs to choose from and they're already written for you.
In particular, skilled people have worked hard to write standards-
compliant serialiser methods on them, including support for varying
character encodings. These things work. They work better than most
people have the skill to duplicate. They take more time to write again
than the competent people can afford.

It's like cryptography. There's only half-a-dozen people who should
ever be allowed to write the components, one of them's insane, one's
Finnish, two are academics with scary hair, and the others are kept in
a locked cupboard by the NSA. The rest of us should only ever re-use
these components, not re-invent them.

So your serialiser should understand the difference between SGML and
XML, or at least HTML and XML. Then you tell it which, and it just
works. If it doesn't, it's broken (so why trust it to get anything
right?).

If you aren't using some sort of intermediate DOM (i.e. direct
document.write()s) then fix that first. _Especially_ for anything
resembling XML.


Here's a spare Clue: If you come to my project again and tell me that
"We need to write a new serialiser from scratch, because the standard
one doesn't work because "Our Project Am Spesshull.", then you'll get
a swift dose of Clueiron justice. I've had this imposed on me three
times now, all incorrectly. First time was dumb, second time was dumb,
a big project and disastrous, third time I just blew the bastard thing
clean out of the source repository (and we all lived happily ever
afterwards).

Mostly people think this in the first place because they don't grok
character encodings.

[Back to original message]