> From rdebry at us.ibm.com Fri May 30 14:08:44 1997
>> RKD> I am really hung up on two issues:
> RKD> The first is just the comprehension issue, which to some
> RKD> extent can be fixed as suggested above.
> RKD> The second issue is one that I caught when re-reading
> RKD> the document for the nth time and thinking about the
> RKD> code problem in dealing with different attribute names
> RKD> and trying to correlate them. That's why going back
> RKD> (atleast on the wire) to a flag or an adornment makes
> RKD> better sense to me. It would make the client code much
> RKD> easier to write and much less prone to errors if the
> RKD> notion of an adornment or flag to the base attributet
> RKD> were used.
In an earlier email you cited the sentence in the model document (lines
1368-1379) "there is no general rule for associating "xxx" with
"xxx-supported" ...". At the San Diego meeting we decided that keyword
variants should be built in a regular fashion even if the rules for
English were violated. We wanted xxx and xxx-supported always to be
related. For example, it is 'media' and 'media-supported' even if it should be
'medium' by the rules of English. This gives us the adornments that
you want without putting them in the value.
At that meeting we left undetermined whether the variant should be
denoted by prefixes or suffixes, though currently we are using suffixes.