Jason Kottke writes:
a distributed data storage system would take the place of a local storage system. And not just data storage, but data processing/filtering/formatting. Taking the weblog example to the extreme, you could use TypePad to write a weblog entry; Flickr to store your photos; store some mp3s (for an mp3 blog) on your ISP-hosted shell account; your events calendar on Upcoming; use iCal to update your personal calendar (which is then stored on your .Mac account); use GMail for email; use TypeKey or Flickr’s authentication system to handle identity; outsource your storage/backups to Google or Akamai; you let Feedburner “listen” for new content from all those sources, transform/aggregate/filter it all, and publish it to your Web space; and you manage all this on the Web at each individual Web site or with a Watson-ish desktop client.
Think of it like Unix…small pieces loosely joined. Each specific service handles what it’s good at. Gmail for mail, iCal for calendars, TypePad for short bits of text, etc. Web client, desktop client, it doesn’t much matter…whatever the user is most comfortable with. Then you just (just! ha!) pipe all these together however you want with services (or desktop apps) handling any filtering/processing that you need, and output it to the file/device/service of your choice. New services can be inserted into the process as they become available.
What you need is a “virtual desktop.”