Fortunately the volume shifts focus completely at this point. Michael Benedikt's own "Cyberspace: Some Proposals" is the centrepiece of the collection, running to over a hundred pages. In it he suggests some formal principles to be used in the construction of virtual "spaces" suitable for information representation. Benedikt occasionally indulges in flights of fancy (cyberspace may provide insights into cosmology?!) and he goes a little overboard with the mathematical formalism (which does makes a pleasant change from the waffliness of the preceding works), but his ideas have obvious connections to real systems. The second part of "Cyberspace: Some Proposals" looks at some proposals for cyberspaces built using these guidelines, complete with colour illustrations. These are speculative, but still solidly rooted in practical problems. More than anything else I have read, this work has convinced me that "cyberspace" can be a useful term and that it doesn't have to be a catch-all for any and everything trendy to do with computers.
Marcos Novak's "Liquid Architectures in Cyberspace" shares Benedikt's architectural perspective, but is more poetical than mathematical: "Cyberspace stands to thought as flight stands to crawling. ... Cyberspace is poetry inhabited, and to navigate through it is to become a leaf on the wind of a dream". Alan Wexelblat's "Giving Meaning to Place" waves the word "semantic" around a lot ("semantic dimensions" and "semantic spaces"), but is basically a fairly boring rehash of basic data structures (I can't help thinking that relabeling discrete variables "quantum dimensions" is a pointless exercise). Carl Tollander's "Collaborative Engines for Multiparticipant Cyberspaces" tries to apply ideas from Edelman's "neural Darwinism" to cyberspace systems, while Tim McFadden's "Notes on the Structure of Cyberspace and the Ballistic Actors Model" uses yet another formal model. Both papers contain some interesting ideas, but in both the theoretical framework employed seems far more detailed than is warranted by the speculative nature of the systems involved. McFadden also has problems reconciling his formalism with his emphasis on what he calls the books (which this time include works by Rudy Rucker as well as Gibson), and his grasp of the philosophical issues involved seems dubious (a claim that the "program of reductionism" must be successful for Gibson style cyberspaces is just bizarre).
I found "The Lessons of Lucasfilm's Habitat" (Chip Morningstar and F. Randall Farmer) really interesting: it goes into some detail about the design and operation of the Habitat system, about which I knew very little. This material is perhaps mostly of historical interest now, since the lessons in question have really been hammered home by years of experience with MUDs. Meredith Bricken's "Virtual Worlds: No Interface to Design", about "total immersion" cyberspace systems, also draws on experience with real implementations.
"Corporate Virtual Workspace" (Steve Pruitt and Tom Barret) and "Making Reality a Cyberspace" (Wendy Kellog, John Carroll, and John Richards) present descriptions of cyberspace systems which are beyond current capabilities but which are clearly becoming plausible. The first paints a rosy picture of future corporate workplaces (it reads a bit like a sales brochure, actually!), while the second considers augmenting reality to give it cyberspace properties as an alternative to creating separate simulated environments.
While many of the works in Cyberspace: First Steps are now a little dated (and some of them are best forgotten), it is still an important collection. It is worth a look by anyone involved with either the information representations or the interfaces of virtual reality systems.