TEI Syntax and Metalanguage Committee
 
                     Exchanging Documents
 
                   Document Number: TEI MLW24
 
                         July17, 1990
 
In the TEI P1 guidelines, on page 2, the following conditions are set
out for document interchange:
 
 1. adherence to the TEI SGML declaration and DTD
 2. documentation for all tags not in the DTD
 3. inclusion of the TEI header at the beginning of the document
 4. all text must be in the given subset of the ISO 646 character
encoding
 
Additionally, some way should be found of ensuring that all the pieces
of the document are present. That is, all entity references are
expandable upon receipt of the document, except those that are public
and are therefore already at all sites. If a file of entity definitions
is missing or if the author forgot to include an entity definition, the
document will not be completely readable. Especially if the guidelines
given in chapter 8 of TEI P1 are followed and the number of entities is
large.
 
One solution to this problem is the SGML Document Interchange Format
(SDIF) which is an ISO standard (9069) meant for this purpose. The
standard calls for a packer/unpacker pair which takes a set of one or
more related documents and marks each entity reference in them. The
document package is wrapped in some SDIF identifiers and all entity
references are checked. The resulting data stream then can be unpacked
upon receipt, with a reasonable guarantee that all entities referenced
will be defined. The details of the packing and unpacking are contained
in the standard and are unimportant here.
 
Since TEI requires something like SDIF to ensure that whole documents
are interchanged, it would be nice if TEI could adopt SDIF immediately.
Unfortunately, in our investigation of this standard, we have not found
an implementation of SDIF, nor anyone who has actually used it. It
seems to meet or exceed the requirements, but it is not really usable
as yet. Perhaps if a TEI document packer is implemented which does such
things as putting the document in canonical form and ensuring that its
character coding set conforms to the ISO 646 subset, it could also use
SDIF to ensure that entity definitions are present or part of some
public set of entity definitions. Alternatively, putting the document
in a canonical form could check the entity references as well, negating
the need for SDIF. Since SDIF is not being used by, seemingly, anyone,
TEI can set its own course without risk of damaging the universatility
of its document interchange.