Found an an early draft of some concept materials for the original MetaWrap Project “Fish” (1998-2001) that could turn any existing HTML website into an XML data-source by the application of the MetaWrap pattern language.
This single XML file recursively defined a set of XML pages that were scraped from the existing website. These XML pages could be dynamically transcoded into any target platform that had a supported plugin. Target platforms were XML, HTML, WAP (over a hundred mobile phone profiles) and OpenTV were supported.
Created GUI tools for selecting sections of the website. The last version we created enabled you to mark sections off in internet explorer and automatically create the appropriate pattern. Was well on the way to creating a tool that could analyse site structure and automatically generate a robust pattern before “The Accident” blunted the worlds desire for big wacky ideas 🙂
Click here then click on choose the “Developer Demo”. This demo is of the first generation of the GUI and its rather rough, but it gets the point across.
James McParlane's Blog 
aaah, fish. now that brings back some memories!
got anything from Octopus, James? 🙂
(no, seriously)
i’d been trying to block the memories *out*!
I have some old octopus demos somewhere 🙂
I still think that stuff is cool, well ahead of its time.
Its nice to see some of my estranged web design skills in the screenshots 🙂
Pattern matching is such a hard problem to solve though, i still think im doing things the wrong way on a daily basis. Like you said, the web is all javascript and absolute positioning now, reliable structure is a thing of the past.
Check out the Aardvark< extension for firefox, its uses a great strategy for selecting page elements.
http://extensionroom.mozdev.org/more-info/aardvark
Where are the octopus demos? 😉