Return of the Clerver--Opera Unite and Michael Dertouzos

In the mid 1990s at MIT, I spent quite a bit of time with Michael Dertouzos because we were both doing some consulting for Siemens Corporation’s IT division.  Michael was heading up the Laboratory of Computer Science at MIT, which combined with the AI Lab is now MIT CSAIL.  Michael was great at many things that made LCS a big success, including fund raising and attracting and supporting top talent.  He was also a very quirky thinker.  If you want to know just how quirky, ask me some day about his take on bathroom hygiene.

Early on in the growth of the web, Michael was bemoaning the separation between clients and servers.  He was fond of pointing out how powerful most home PCs were already back then and how utilizing them only as web clients was making little use of their capabilities.  He would joke that what he wanted instead was a “clerver” (a client+server).  We wound up getting some of that but only in the peer-to-peer world which is still associated primarily with filesharing (and at least by association with illegal filesharing), whereas the core architecture of the web remained one of comparatively few centralized servers and many clients.  As an aside, Michael was instrumental in bringing Tim Berners-Lee to MIT and establishing the W3C.

Well Michael would be happy about Opera Unite, which was announced today and essentially turns the Opera web browser into a web server also.  At first glance, this is far from revolutionary.  Every Mac with OS X has a version of Apache on it that can be turned on from the Sharing preferences.  But that doesn’t do much good because most home machines are NATed (or “natted”) which means that their web servers are not accessible from the web.  I have not dug in yet how Opera Unite solves this but making lots of machines accessible as bona-fide servers has the potential to enable new and interesting services (not necessariily the ones that Opera is touting, but home energy monitoring and control is one example that comes to mind).

Of course, the right way to solve this problem would be to somehow get our act together and finally make the transition to IPv6.  Then all home machines and other intelligent devices could be on the network directly and could use standard web servers.  Sadly, I can easily see us going another decade or more without making a meaningful transition to IPv6 (this might have been a good use of stimulus funds!).