SGML [features, attributes, relationships]

[Joyce Flaherty <flahertj -at- SMTPGW -dot- LIEBERT -dot- COM>]

Text item: Text_1

Long ago in another life I worked on a mainframe project we
called a data dictionary.  The vendor software we used was
called Datamanager.  The project lasted about six months
during which we identified a set of attributes (for example,
language, compiler, preprocessor, link-edit options, NCAL,
re-entrancy) for every source code item we owned.
Additionally, we identified all relationships for every
software item (system, program, module, macro, clist, proc) we
owned.  The relationships we identified and fed into the
dictionary provided the query function: "what uses X?" and
"what is used by X."  We used the dictionary to measure the
impact of a change, among other things.

It seems to me that an unambiguous, richly-encoded, highly
content-tagged SGML database that apes the data dictionary
would provide the *features* function we want.  If a
customer wants product X with options Y and Z, you retrieve
the doc to support product X AND options Y and Z.  Then you
add to doc entity Y and Z the *used by* attribute.  I'm
still struggling with how much of this information in stored
in the SGML database, and how much is more appropriately
maintained externally.

What do you think, SGML persons?

BTW, there is a PC equivalent tool for the "what is used by
X?" function.  The setup program Wizard, uses it.  I don't have
a handle on how it works yet, but I'm working on it.  Either it
runs your program and captures every call on the fly, or the
calls are stored as part of the compiled EXE as attributes. I
think it's amazing that a subroutine I know almost nothing
about, knows things about my program that I don't know.  I have
installed programs in various environments only to find when
I test the program that I didn't include a required DLL.

joyce flaherty
flahertj -at- smtpgw -dot- liebert -dot- com