A major weakness of the HTTP protocol is its inability to share a single connection between multiple simultaneous transactions — to multiplex a number of transactions over a single connection. In HTTP, a client can either request all instances sequentially, which significantly increases latency, or else open multiple concurrent connections, with all the problems that this implies (see Persistent connections).
Poor Man’s Multiplexing (PMM) is a technique that simulates multiplexing by requesting an instance in multiple segments; because the segments are fetched in independent transactions, they can be interleaved with requests for other resources.
Obviously, PMM only makes sense in the presence of persistent connections; additionally, it is only effective in the presence of pipelining (see Pipelining).
PMM poses a number of reliability issues. If the resource being fetched is dynamic, it is quite possible that it will change between segments; thus, an implementation making use of PMM needs to be able to switch to full-resource retrieval when it detects a dynamic resource.
Polipo supports PMM, but it is disabled it by default (see PMM).