Re: ISSUE-54: doctype-legacy-compat

Lachlan Hunt wrote on 01/16/2009 10:03:17 AM:
>
> Henri Sivonen wrote:
> > <!DOCTYPE html SYSTEM "about:sgml-compat"> would seem to fit the
> > requirements as a legacy generator alternative to <!DOCTYPE html>.
>
> Based on some further discussion in IRC, these are the advantages of
> using the SYSTEM identifier with an about: URI scheme.
>
> * Many people are already familiar with the scheme and so it's
>    memorable.  More so than, e.g., the tag: or data: URI schemes.
>
> * It's shorter than an equivalent http: URI would be
>
> * It's non-retrievable and so there's no expectation of a DTD.
>
> * Use of an absolute URI rather than a string like "sgml-compat" that is
>    indistinguishable from a relative URI prevents consumers that attempt
>    to retrieve DTDs from getting 404s from many servers.
>
> * Typing about:sgml-compat into browsers generally results in an error
>    message or blank page. This helps in pre-empting any bogus
>    rationalisations for why using this one with the URI is better than
>    <!DOCTYPE html> without the URI.
>
> * As Philip demonstrated earlier, using a SYSTEM identifer is generally
>    more compatible with legacy producers than a lone PUBLIC identifier.

Excellent!  Is this something everybody can live with?  If not, please
speak up.

As I tally it so far, there are strong objections to PUBLIC "" and PUBLIC
"XSLT-compat".  The objections to "legacy-compat" aren't quite as strong
and I haven't yet seen somebody say that they could not live with that.  As
I see it, the this proposal doesn't match any of the previously raised
objections, and would be preferred over legacy, provided:

1) it were optional, and that the shorter version is still considered
conformant.

2) if there are any recommendations in the spec, they don't single out any
specific tool.

- Sam Ruby

Received on Friday, 16 January 2009 15:22:34 UTC