- From: Henry S. Thompson <ht@cogsci.ed.ac.uk>
- Date: Thu, 7 Nov 96 12:44:37 GMT
- To: Arjun Ray <aray@nmds.com>, W3C-SGML-WG@w3.org
On the general subject of empty elements, DTD info reprised in <?XML ... ?> PIs, etc, I find myself in a state of serious uncertainty. On the one hand, I REALLY care about SGML forward compatibility, so I'd be REALLY sorry to lose vanilla <e> empty elements. And I already work with a normalised form very close to XML as the draft DID have it, using a concise DTD reprise, so I could see converting my existing code to handle XML plus DTD-reprise-in-PIs very easily. On the other hand, I'm sympathetic to the criticism of adhocery which Arjun makes, and indeed made it myself in the debate over character-set-declarations-as-PIs. I must say I agree with Eve Maler that grandfathering a specific LIST of empty elements is reprehensible. Note that wrt the 'can a perl hacker do it in an afternoon' issue, this makes the published decision virtually indistinguishable in difficulty from the DTD-reprise-in-PIs approach---in either case a table of empty GIs has to be maintained and checked. I also agree with Paul that the 'possible inconsistencies' argument is vacuous. On balance, then, I go with my original response on reading the draft, namely "Oh good, a serviceable approach to compatibility", even if it means using PIs IFF YOU WANT TO USE <e>. Oh yes, one final note, wrt Arjun's remarks about PIs: a parser IS an application, and there's nothing bogus about shortstopping PIs flagged for the parser itself, so the claimed necessity of having to round-trip the PIs out and back again is invalid and not as such a basis for rejecting this approach. I too therefore urge the board to return to the status quo ante wrt <e> and <?XML empty names ... ?> ht
Received on Thursday, 7 November 1996 07:45:29 UTC