Конвейеризация HTTP: история, проблемы и замена современными технологиями.

HTTP pipelining is a feature of HTTP/1.1, which allows multiple HTTP requests to be sent over a single TCP connection without waiting for the corresponding responses. HTTP/1.1 requires servers to respond to pipelined requests correctly, with non pipelined but valid responses even if server does not support HTTP pipelining. Despite this requirement, many legacy HTTP/1.1 servers do not support pipelining correctly, forcing most HTTP clients to not use HTTP pipelining. The technique was superseded by multiplexing via HTTP/2, which is supported by most modern browsers. In HTTP/3, multiplexing is accomplished via QUIC which replaces TCP. This further reduces loading time, as there is no head of line blocking even if some packets are lost.

The pipelining of requests results in a dramatic improvement in the loading times of HTML pages, especially over high latency connections such as satellite Internet connections. The speedup is less apparent on broadband connections, as the limitation of HTTP 1.1 still applies: the server must send its responses in the same order that the requests were received—so the entire connection remains first in first out By 2017 most browsers supported HTTP/2 by default which uses multiplexing instead. Read requests like GET and HEAD can always be pipelined. A sequence of other idempotent requests like PUT and DELETE can be pipelined or not depending on whether requests in the sequence depend on the effect of others. Most pipelining problems happen in HTTP intermediate nodes (hop by hop), i. e. in proxy servers, especially in transparent proxy servers (if one of them along the HTTP chain does not handle pipelined requests properly then nothing works as it should). The same advantage exists in HTTP/2 multiplexed streams.

Pipelining was introduced in HTTP/1.1 and was not present in HTTP/1.0. There have always been complaints about browsers, proxy servers, etc. not working well when using pipelined requests / responses, up to the point that for many years (at least till 2011) software developers, engineers, web experts, etc. tried to summarize the various kind of problems they noted, to fix things and to give advices about how to deal with pipelining on the Open Web.

Of all the major browsers, only Opera had a fully working implementation that was enabled by default. In other browsers HTTP pipelining was disabled or not implemented. Internet Explorer 11 does not support pipelining. Mozilla browsers (such as Mozilla Firefox, SeaMonkey and Camino) used to support pipelining; however, it was removed in Firefox 54. When it was supported, pipelining was disabled by default to avoid issues with misbehaving servers. If pipelining was enabled by the user, Mozilla browsers used some heuristics, mostly to turn pipelining off for older Microsoft IIS servers. The removal was eventually backported to SeaMonkey. Konqueror 2.0 supports pipelining, but it is disabled by default. Google Chrome previously supported pipelining, but it has been disabled due to bugs and problems with poorly behaving servers. Pale Moon (web browser) supports pipelining, and is enabled by default.

Most HTTP proxies do not pipeline outgoing requests. Some HTTP proxies, including transparent HTTP proxies, may manage pipelined requests very badly (e. g. by mixing up the order of pipelined responses). Some versions of the Squid web proxy will pipeline up to two outgoing requests. This functionality has been disabled by default and needs to be manually enabled for "bandwidth management and access logging reasons". Squid supports multiple requests from clients. The Polipo proxy pipelines outgoing requests. Tempesta FW, an open source application delivery controller, also pipelines requests to backend servers.