A recurring theme on forums like flexcoders is the issue of user interface architecture within Flex applications. Rich-client applications developed using frameworks like Flex are something of a departure from ‘traditional’ web applications, so the architectures employed for such applications will inevitably differ. But not so long ago we were comparing the new concept of web-application development with that of ‘traditional’ rich-client applications and arriving at very similar conclusions. We have come full circle, from rich-clients to web-clients and back again. So what lessons can we learn from the 20 or so years of rich-client development preceding Web 1.0? And are there web-application patterns that translate across to the rich-client world?