xml-dev - Re: User passed Arguments for Entities

[Rick JELLIFFE <ricko@geotempo.com>]

Toby Speight wrote:
> 
> 0> In article <3A0F8A28.D96A5310@geotempo.com>,
> 0> Rick JELLIFFE <URL:mailto:ricko@geotempo.com> ("Rick") wrote:
> 
> Rick> In SGML, you have a richer variety of entity declarations
> Rick> mechanisms that might allow you to do something.  In SGML, you
> Rick> would probably rewrite the storage manager to allow reading
> Rick> entities from standard input; some SGML systems allowed this.
> Rick> You could hack the storage manager for your XML processor,
> Rick> make up some home-made convention for accessing standard input
> Rick> etc.
> 
> No hacking or home-made conventions required if you're using the SP
> toolset (which handles XML as easily as any other kind of SGML).  A
> formal system identifier (FSI) for standard input looks like "<osfd>0".

Oops, I should have remembered them!

Formal System Identifiers were one of the additions to SGML in the
mid-90s. For layering they were grouped in the SGML General Facilities
and parked in with the HyTime spec pending the revision.

FSIs allow a psuedo-start-tag inside a SYSTEM identifier, including with
psuedo-attributes.  There are many defined FSIs, for the major different
types of storage systems.  What was very interesting about them was that
one can pipe them.  For example, you could have an entity defined to be
a ZIP file, then another entity defined to be the tar file resulting
from unzipping the first entity, then another entity defined as being a
certain file extracted from tar -xf of the second entity.  

When XML came along, it was decided to use URLs as the system
identifiers, so FSIs are not really available (AFAIK).  

This is an example of how SGML attempts to solve different problems than
XML. SGML is much more concerned with archiving, management and
construction of documents; XML is more concerned with distribution and
unaided processing.

Cheers
Rick Jelliffe