The May 05 edition of IEEE DSOnline includes a very interesting survey on research within
System architectures for Web content adaptation services.
In the article,
Web content adaptation means transcoding, filtering or personalizing Web content somewhere between the content provider and the receiver.
Basically, transcoding brings content from one format to another when the receiving device does not support the original format, or network or computing resources require a smaller format.
Personalization means using metadata to enhance the content, increase relevance or remove unwanted content, and may include transcoding.
The authors present three different architectural choices; client-side, server-side and (inter-)mediary-adaptation, and discusses the weaknesses and strenghts of the different approaches, as well pointing out "missing links" in the area.
I especially found it interesting that they pointed out how state-ful personalization is very hard for mediators. Their argument is that in a mediator network, a client-server session is not expected to use the same mediator twice, and hence keeping transaction traces demands some shared storage, which brings us back to the centralized solution again. Or, we could take the WAIF-approach, and make sure mediators keep state, and that clients use the same mediator every time.
In sum, the article provides a fine overview of one of the facets of the WAIF project, and many links to WAIF-related stuff. Hereby recommended.