- From: Thomas Broyer <t.broyer@ltgt.net>
- Date: Fri, 16 Jan 2009 16:04:55 +0100
- To: public-html <public-html@w3.org>
- Message-ID: <a9699fd20901160704h52d5fcb3q967514baad2e9d3@mail.gmail.com>
On Fri, Jan 16, 2009 at 3:11 PM, Robin Berjon wrote: > > On Jan 16, 2009, at 14:48 , Sam Ruby wrote: >> >> One such tool is mentioned in the bug report[8] cited in my prior email: >> >> [8] http://www.w3.org/Bugs/Public/show_bug.cgi?id=6336 >> > > As one who uses such tools, I would tend to think that I would expect them > to require updates to their HTML serialisations in order to support new > empty elements anyway, and therefore that tossing in a change for the > DOCTYPE wouldn't be much work. For the user who really wants to produce > HTML5 the tooling update is minimal; and if there's anyone out there who is > at the same time so cutting edge that they want to produce HTML5 but so > conservative that they won't upgrade a serialisation library I would tend to > think that they have enough issues of their own that we don't need to > meddle. And for them, an easy workaround is to prepend a hard-coded <!DOCTYPE html> at the end of their serialization chain (in Java, that would mean plugging the attached Writer –which could probably be improved performance-wise– somewhere between your producer/serializer and output stream). So I totally agree with you that it's not really a problem... ...except for tools based on specs that do not allow outputting the HTML5 doctype (such as XSLT): updating the tool is not enough, the spec has to change too; and if it already has been superseded (such as XSLT 1.0) there little to no chance that it'll be updated (and that then tools align with the updated spec). So I think a "compat" DOCTYPE has to be allowed. I'm fine with either: <!DOCTYPE html PUBLIC "legacy-generator-compat" ""> (provided noone can point to a tool that couldn't output an empty system ID) or <!DOCTYPE html PUBLIC "legacy-generator-compat" "about:whatever"> or <!DOCTYPE html SYSTEM "about:whatever"> where about:whatever is any about:* URL (or data:* URL? or tag:* URL? or actually any URL). This I don't mind, but I'm opposed to <!DOCTYPE html PUBLIC ""> and similar proposals for the reasons already expressed by others (error-prone, not evident why it exists so will lead to flamewars, etc.). -- Thomas Broyer
Attachments
- application/octet-stream attachment: Html5DoctypePrepender.java
Received on Friday, 16 January 2009 15:05:31 UTC