Re: ISSUE-54 (html5-doctype-vs-xslt): XSLT 1.0 can not generate HTML5 documents [HTML 5 spec]

Jirka Kosek <>, 2008-08-29 10:17 +0200:

> Henri Sivonen wrote:
> > I'd be OK with <!DOCTYPE html PUBLIC "XSLT-compat">, since it reflects
> > the problem it is solving--making the string resistant to bogus
> > rationalizations about its purpose.
> That's indeed way to go. As HTML5 is not an SGML application there is no
> reason to stick to FPI syntax inside "public identifier". I think that
> this proposal will accommodate all parties.

A couple of questions/comments -

  - Instead of "XSLT-compat" or some other arbitrary string, why
    not just require (if we do decide we want to allow it
    there at all) that it just be the empty string? XSLT engines
    can output an empty value for it, and it seems to me that if
    we have it all, we would want the value to be empty, not some
    standard value that would become a de facto public ID and that
    apps would lead to the same very real "bogus rationalizations
    about its purpose" problem that Henri describes.

  - I don't know how to get xsltproc to output just
    <!DOCTYPE html PUBLIC "XSLT-compat"> (or with whatever string
    for the public ID).

    If I do xsl:output@doctype-public="XSLT-compat", xsltproc
    gives me:

      <!DOCTYPE html PUBLIC "XSLT-compat" "">

    That is, with an empty string for the system ID. If I instead
    do xsl:output@doctype-system="XSLT-compat", I get:

      <!DOCTYPE html SYSTEM "">

    So unless there's something I'm missing about how to get
    xsltproc to omit the system ID when specifying a public ID,
    the above seems to suggest that if we go with this proposal at
    all, it perhaps out to be that it be that it be to the doctype
    to optionally include the SYSTEM keyword plus "" or ''.

Michael(tm) Smith

Received on Friday, 29 August 2008 09:04:47 UTC