- From: len bullard <cbullard@hiwaay.net>
- Date: Fri, 16 May 1997 19:27:46 -0500
- To: Arjun Ray <aray@q2.net>, w3c-sgml-wg@w3.org
len bullard wrote: > > Arjun Ray wrote: > > > > On Fri, 16 May 1997, Dave Hollander wrote: > > > > > I am shocked; there have been no postings about this proposal. > > > > 8 months ago, I was one of the few who argued for empty end-tags. The ERB > > decided otherwise. > > > > But I agree, the lack of posts on this proposal (even to heap scorn on > > it:-)) is somewhat surprising. > > > > Arjun > > Some decisions for simplicity were probably not simplifying. > > The use of the empty end tag is well established in IDE/AS > and used a lot when entering manually and converting. Without > pretty print indentions, it does make reading a little harder, > and for someone doing simple macros based on search and > replace, empty end tags are less than idea. However, since > the specification does not mandate them, having that option > is attractive and one less thing to change in existing code. There is one other argument for them. In some SGML systems, object-likeness was achieved by using the GI as a post-processed object class identifier: <!ELEMENT GeorgeISSonOfSam silly example but also not the point. By using this orthogonal processing, it was possible to create database effects. Now, while this may be a less than ideal solution, that is also not the point. The problem was these extended GIs when required to have matching end tags raised the file size considerably. It is a problem for any design that uses the GI as basis for a compound morphology. Whether this issue is a priority one issue or not is a coin toss. For my money, it isn't a high priority but a nicety. OTH, I'm used to it so maybe have a higher comfort level. Some systems already do it, and others make it an option. While the arguments for readability and desperate perl hackers are sympathetically met, the need for optionality seems clear in that IMO, it is the responsibility of the application designer to decide that an endtag must be *named*. Only the application designer knows about the processes and consumers of the output of processes. len
Received on Friday, 16 May 1997 20:28:04 UTC