Projects

This is Steve Savitzky's personal list of things that ``it would be nice if'' somebody could do.

Agents and Tagsets in XML

This is in progress. See Design/xml-world.html

Document Representation

This is in progress. See Design/dom.html

Cleanup

XML-isms

This has been done, except for DTD's.

XML is a lot more restrictive than SGML and HTML. All attribute values must be quoted, for example, and all attributes must have explicit values. All tags must be closed.

There are really several related projects here:

• Fix the documentation to make it clear that (our version of) XHTML is not HTML with XML syntax, but an extension of HTML with SGML syntax. Probably, two different tagsets will be required, and two different filename extensions.
• Make sure that the what reaches the browser is acceptable XHTML when called for. Unfortunately the Web is not moving in the direction of SGML, but away from it toward XML.
• Of course, we still need DTD's. There needs to be a tagset that transforms a tagset file into a DTD, and tagsets will have to be extended in order to make this possible.

Ports

These are covered in more detail in porting.

C

The server portion of a C port could be built on top of Apache. Certainly the DPS could easily be made into a module. Adding Agent functionality to Apache would be somewhat harder.

In any case, an Apache ``modDPS'' would be a worthwhile thing to have.

Scheme/Guile/DSSSL

The standard SGML style language DSSSL is basically a set of functions built on top of Scheme. SGML parsers exist, obviously. A DSSSL version of the DPS should be trivial.

Parsers

Most SGML parsers in the ``real world'' operate in ``push'' mode -- they call an action routine whenever a tag or text node is encountered. The PIA, in contrast, operates in ``pull'' mode, requesting nodes from the parser.

In order to interface a standard parser to the PIA (which would be a good idea for many reasons) one would have to do one of the following:

• generate a complete parse tree (as a DOM Document) and traverse it with an Input. Trivial, but requires the entire document to be in memory.
• use two threads, with a queue holding nodes that have not yet been traversed by the DPS.
• Revise the DPS to operate entirely in push mode. The old Interform interpretor could do this; it involves building partial parse trees for the content of active tags.

Alternate Document Representations

Object Database

By putting DOM objects in an ODBMS, one could totally replace the filesystem. One would, of course, still need a parser for incoming HTML.

CORBA DOM

Assuming a CORBA-compatible DOM implementation existed (since it's defined in IDL it probably wouldn't be hard), one could use it for document storage.