Date: Mon, 22 Jan 90 15:17:12 EST
From: amsler@flash.bellcore.com (Robert A Amsler)
Message-Id: <9001222017.AA29252@flash.bellcore.com>
To: ide%vassar.bitnet@cunyvm.cuny.edu, susan%vax.ox.ac.uk@nss.cs.ucl.ac.uk,
        u18189%uicvm.bitnet@cunyvm.cuny.edu, walker@flash.bellcore.com
Subject: Gary Simon's draft - some thoughts

(Both from the perspective of Metalanguage and the Dictionary Standard).
I'm preparing my comments and think I've figured out what the problem
is. It involves the difficulty of maintaining the order of the original
information in a text and recording the scope of the information.
It would seem that if one moves information to where it has to be
to show the scope within the strictly hierarchical structure of SGML
one loses the ability to reconstruct the original order of the information
in the text. I did this in the dictionary standard, but since
I wasn't particularly worried about reconstructing the actual printed
representation I didn't notice the real problem. Printed works often
have multiple locations at which information can appear. Since all scoping
for them is carried by semantic or pragmatic information, this doesn't
bother them--but in SGML the scoping is solely carried by the nesting
of tags.

If  instead  one  leaves  the  information  where  it  appears in the
original text,  as was  done in  the OED2,  one has  no mechanism for
recording the scope of the tagged items.  There is no  problem if all
items  appear  at  the  point  at  which  their scope  starts, but in
dictionaries at least, that is not true.  The issue of attributes vs.
tags  may  have  obscured  the  actual  problem  of  recording  scope
information for attributive content.

Anyway... that is how I currently see things...