Ron Lake's The Architecture of the GeoWeb mishandles a subtle point of the Web's architecture. Web crawlers don't find and read files, they traverse the web of resources, URL to URL, and parse the documents that represent those resources. The contents of enterprise spatial databases are not passed over because they are not files, but because they have no URLs. They are not of the Web.
Any Rails, Django, or Zope (for example) application has resources that are stored in a database, and these are duly indexed because they have URLs. The geographic places of Pleiades, such as Antiphellos/Habesos, persist in an object database (ZODB) and have URLs. These places are connected to the Web by virtue of their links from other documents, and are spatially referenced by linking to KML or GeoRSS representations. This is a geo-web architecture that truly builds upon that of the Web.
Go read Lake's post. It's a good overview of new web spatial indexes, and the current isolation of enterprise geo-data from the Web.