I'd really like a keylogger for OS X that archives not only keys pressed but names of the applications they're typed in, exporting to monthly files. I'd like daily automatic HTML dumps of my Firefox browsing history. If there were large amounts of storage available, I'd like some kind of movie (preferably retaining the text as text, when text is rendered, like an SVG movie of sorts) of my entire computing interactions.
I'd like all of the books that are in the public domain to be freely available as images and OCR'd text documents online. Google Books for some reason doesn't automatically place books written before c.1930 into its Full View category, though I believe that legally they could do so. Manuscripts should also be brought into the fray, especially when OCR becomes palæographically competent.
I'd like all photographs on Flickr to be linked together just like Microsoft are doing with their Photosynth technology, but I'd also like to be able to apply this to my own photographs and create interactive 3D models of the landscapes that I take photographs of. It'd be nice if this could be linked into satellite imagery too, as well as the Flickr environment.
I'd like all train and bus and other schedules to be online in a good machine readable format, JSON most likely being the best choice, so that people can compete to produce excellent accessible user interfaces for them. Why should we have to use a single company interface? It's in the interests of these kinds of companies to provide this data to their users, so that their services are used more.
I'd like the government to give subsidies to the Ordnance Survey so that all map data is free. Same for the Oxford English Dictionary: the magic oed.sgml file should be in the public domain. If not tax pounds, then perhaps a donation or a research council fund? It's antithetical to the information age to have all of this exceptional data locked up behind pay barriers.
Google Maps and Flickr Maps are great, but I'd like more integration of the online and offline worlds geographically speaking. It should be possible to annotate the landscape more easily and pick that up whilst you're walking around; William Loughborough calls this “ambient information”. There are three ways in which this might be achieved: more central sites that people can collaborate on à la Wikipedia; more geotagging of content; or more Thought Treasure-like NLP extraction processes.
I'd like computer systems to be more analogue, so that they don't lose the splendour of Dickinson's dashes. In other words, to support more gaiji for example, and to be less reliant on decades old datastructures and techniques. I should be able to embed little SVG sketches in my pages more easily, and they should scale along with the text.
I'd like to be able to better manage my personal digital library. I probably have around a terabyte or two of data, but because it expands at very close to the rate of available storage, it's strewn across several systems and discs and in lots of different organisational styles and formats. It's exceptionally fractal, and requires not only elbow grease but some techniques to help manage it effectively.
The Semantic Web isn't concerned with any of this, so I'm going to stop working on the Semantic Web. By this, I mean anything to do with RDF, OWL, Dublin Core, FOAF, SIOC, and so on. I will no longer be writing or supporting Trio, my RDF API. I will no longer give answers, help, or advice to people wanting information about Semantic Web technologies. I will, moreover, actively dissuade anyone from working on the Semantic Web where it distracts them from working on the use cases outlined herein.