Using the Client Cache for Content Encoding: Shared Dictionary Compression for the Web

As different approaches have demonstrated in the past, delta encoding and shared dictionary compression can significantly reduce the payload of websites. However, choosing a good dictionary or delta source is still a challenge and has kept delta encoding from becoming practically relevant for today’s web. In this work, we demonstrate that the often prohibitive costs of dictionary generation exhibited by earlier approaches can be avoided by simply using cache entries for content encoding: We divide web pages into different page types and use one actual page of every type as a dictionary to encode pages of the same type. In an experimental evaluation, we show that our approach outperforms current industry standards by a factor of 5 in terms of compression ratio. We discuss optimization and content normalization strategies as well as application scenarios that are possible with our approach, but infeasible with the current state of the art.