Some people are afraid that Google App Engine won’t be a really big success because it’s a lock-in. Give it a month and no doubt there will be a project which allows you to run AppEngine applications on your own servers, which (in my opinion) is even more interesting than AppEngine itself.
Mostly that won’t be a problem, the challenge would be BigTable, but I guess CouchDB could be used for that.
The question is, would this really be more interesting? I doubt it. People don’t use AppEngine because of it’s great APIs. Let’s face it, it’s more difficult to use than “normal” APIs like httplib, mail libraries and normal interfaces to relational databases. It’s its scalability that makes it interesting. Even if you would get the software from them, you would still need thousands of servers worldwide to scale like they do.
It is already running on Amazon EC2:
http://jchris.mfdz.com/code/2008/4/announcing_appdrop_com__host_go
CouchDB is not quite ready yet anytime soon. I’d place my bets on Hyper Table or Hadoop’s HBase which are explicitly modeled after Google’s Bigtable.
The hurdle with fetching pages and mail isn’t really that big: it’s not exactly what you do all the time you write an app. Bigtable isn’t that bad either, those unfashionable JOINS were going away anyway. And for the common case Django’s abstraction still works fine.
Though the limitations can indeed be a pain, it isn’t just the scaling that is cool about AppEngine. I think that Python will really get a boost as a web application development language. We’ll, no doubt, see a lot of new libraries and frameworks (all with AppEngine like restrictions in mind) written in Python, which is interesting.