[Ed Summers said]

> Michael -- there is a nice side effect of running your SMGL through James
> Clark's SX (the SGML->XML converter that you mentioned
>, in that external entitites are expanded using a
> catalog.  The resulting XML file isn't exactly pretty since it has oddly
> placed line breaks--but (of course) it is valid XML!

Yes, SX is very useful.... but if I recall correctly, what SX does is
generate a series of entity declarations in the internal subset,
complete with system identifiers taken from the SGML Open catalog.

I think one of Michael's points was that XSLT doesn't provide a
mechanism to _access_ the value of that system identifier during a
transformation, so using XSLT, one can't transform from an EAD
element with an entityref attribute pointing to an unparsed external
entity, like

<extref entityref="myentity">

to an HTML

<a href="">

even if SX _has_  generously provided
"" as a system
identifier in an entity declaration for myentity.

What would be nice is the XSLT equivalent of DSSSL's
"entity-system-id" and "entity-generated-system-id" procedures, but I
asked the question on the XSL-List a while ago and got the response
that there wasn't such functionality  - well, actually, it was "no,
but it's an outstanding issue (but it might be difficult to get
support for it)". [I assume that meant support from vendors.] That
was with reference to the previous XSL-WD, but I notice it still
remains an "outstanding issue" in the new XSLT-WD.

Given the DSSSL capability, it would be possible to perform the
SGML>XML conversion using a DSSSL stylesheet which transforms an SGML
document (with entityref attributes) to an XML document (with href
attributes whose values are obtained from a catalog).

This was one reason I have been holding onto the view of the SGML
document as the "original" - given the current problems of
URL-persistence, I do prefer the idea of an entityref-based/href-less
document as an archival version.

However, I acknowledge that it would be neater to be able to get at
this system identifier using XML tools. My lack of programming skills
doesn't enable me to judge, but the new XSLT-WD's (and XT's) support
for "extension functions" suggests a means by which such a program
could be incorporated within an XSLT-based transformation process -
either getting the system id from the entity declaration or going the
whole way and performing a catalog lookup on the public id.

Moving slightly beyond the domain of EAD, I was interested by David
Durand's suggestion that XML's ease of use makes it a better
candidate than SGML for an archival format. As a community with a
particular interest in finding formats with longevity - not just for
our finding aids, but for digital documents in general -, we should
seek support in XML and XML tools for features which enhance that
longevity - particularly if, as David suggests, one of the principal
barriers is that vendors are not being made aware of the demand!

Having said that, of course, I am aware of the danger of asking for
the re-introduction of the very complexity which XML was intended to

Pete Johnston
Pete Johnston (Effective Records Management Project)
Archives & Business Records Centre
University of Glasgow
77-87 Dumbarton Road
Glasgow G11 6PW   E-Mail: [log in to unmask]
Scotland, U.K.    URL:

Tel:  (UK) 0141 339 8855 ext. 0166 or (UK) 0141-330-4159
Fax:  (UK) 0141-330-4158