Interesting to read this from Mitch Kapor, of all people, considering that it was his former company who created one of the best-known replication technologies:
The exception to the far greater convenience of the browser is off-line usage. With no net connection, data stored in a web app is inaccessible. So, infrastructure to support local storage of data (via caching, via something fancier) as a standard affordance of web-based applications is perhaps the biggest remaining barrier to be overcome. There is no fundamental reason I am aware of it can’t be overcome either on a case-by-case basis, or better, in a more general way which would work not just for a given application, but for many of them.