Episode 470: L. Peter Deutsch on the Fallacies of Distributed Computing

Latency Challenges

Latency in network applications presents significant challenges, often underestimated by developers. Peter Deutsch highlights how latency can arise from both network transport and application design, emphasizing the importance of realistic expectations for data transfer 1. He shares a personal experience where combining multiple web resources into a single request improved application responsiveness, illustrating practical strategies to mitigate latency issues 2.

Simply sending all the data in one request actually made a noticeable improvement in the responsiveness of the application.

--- Peter Deutsch

Advancements like HTTP/2 and the QUIC protocol offer hope for reducing latency by enabling multiple requests over a single connection, though current technologies still pose challenges 2.

   

Error Management

Error management in network communication often lacks effectiveness, particularly at the application level. Peter Deutsch argues that while lower-level protocols handle error correction well, application-level outages and interruptions remain poorly managed 3. He criticizes the lack of diagnostic tools and communication about network issues, citing personal frustrations with broadband providers 3.

Handling timeouts, handling failures to transmit is still very important, and software generally deals with that very badly.

--- Peter Deutsch

Deutsch also points out the poor user interface design in applications like email, where failed transmissions are not communicated to users, violating the principle of least surprise 4.