🚀 go-pugleaf

References: <9208242114.AA13720@ucbvax.Berkeley.EDU><23320E@erik.naggum.no>

I wrote:

>Note too that SDIF interchange cannot be reliably implemented
>between systems that have different filesystem structures
>unless they share a common entity declaration convention.
>Such a convention is outside the scope of SGML or SDIF
>definitions and is thus application-dependent.

Both Charles Goldfard offline and Erik N. online made me see the
error of my ways.  The above statement is in fact not strictly
true.

However, it is true to the degree that your non-SGML processes
(file management, human sanity, etc) depend on certain filename
structures and conventions.  To the degree that conventions and
structures need to be preserved, then SDIF is not sufficient without
extra-SGML conventions and processing.  However, that does not in
any way diminish the utility of SDIF, because that problem exists
independently of SDIF, it's just that SDIF doesn't solve it completely.

However, Erik's comments about public identifiers are well taken, and
he's 100% correct.  However, until SGML processing becomes a standard
part of most operating systems, most SGML users will still be working
with real filenames on their local system.  To use IBM as an example,
as we start to build SGML-based systems and processes for managing
information objects, we can start to build into those systems and
processes support for public identifiers so that such support is
consistent and ubiquitous, but that will take time.  In the mean time,
I, Eliot Kimber, am faced with the immediate problem of supporting the
development of SGML-based information across multiple operating
systems with widely differing filesystems using different tools of
varying support for entity management.

My frustration is not with SGML, but with the fact that until support
for SGML is provided ubiquitously, specifically by providing a parser
and standard entity manager on all operating systems, I can't really
take advantage of the full entity management function that SGML
defines and enables.

In analyzing the problems we have within IBM in terms of doing enterprise
publishing of IBM customer documentation using SGML-based systems
and methods, it didn't take long to realize that the big hole is
file management (or more generically, information object management),
and that the only rational solution was one that was built around
a ubiquitous SGML entity manager that provided data access services
for all parts of the system, from authoring tool to downstream batch
processor.  The definition and use of public identifiers will be a key
part of making such a system work.

Eliot Kimber                      Internet:  drmacro@ralvm13.vnet.ibm.com
Dept E14/B500                     IBMMAIL:   USIB2DK9@IBMMAIL
Network Programs Information Development     Phone: 1-919-543-7091
IBM Corporation
Research Triangle Park, NC 27709