pipelining: removed

As previously planned and documented in DEPRECATE.md, all pipelining
code is removed.

Closes #3651
This commit is contained in:
Daniel Stenberg 2019-04-05 16:38:36 +02:00
parent aba1c51553
commit 2f44e94efb
No known key found for this signature in database
GPG key ID: 5CC908FDB71E12C2
28 changed files with 190 additions and 1401 deletions

View file

@ -5,46 +5,6 @@ email the curl-library mailing list as soon as possible and explain to us why
this is a problem for you and how your use case can't be satisfied properly
using a work around.
## HTTP pipelining
HTTP pipelining is badly supported by curl in the sense that we have bugs and
it is a fragile feature without enough tests. Also, when something turns out
to have problems it is really tricky to debug due to the timing sensitivity so
very often enabling debug outputs or similar completely changes the nature of
the behavior and things are not reproducing anymore!
HTTP pipelining was never enabled by default by the large desktop browsers due
to all the issues with it. Both Firefox and Chrome have also dropped
pipelining support entirely since a long time back now. We are in fact over
time becoming more and more lonely in supporting pipelining.
The bad state of HTTP pipelining was a primary driving factor behind HTTP/2
and its multiplexing feature. HTTP/2 multiplexing is truly and really
"pipelining done right". It is way more solid, practical and solves the use
case in a better way with better performance and fewer downsides and problems.
In 2018, pipelining *should* be abandoned and HTTP/2 should be used instead.
### State
In 7.62.0, we will add code that ignores the "enable pipeline" option
setting). The *setopt() function would still return "OK" though so the
application couldn't tell that this is happening.
Users who truly need pipelining from that version will need to modify the code
(ever so slightly) and rebuild.
### Removal
Six months later, in sync with the planned release happen in April 2019,
(might be 7.66.0), assuming no major riots have occurred due to this in the
mean time, we rip out the pipelining code. It is in the order of 1000 lines of
libcurl code.
Left to answer: should the *setopt() function start to return error when these
options are set to be able to tell when they're trying to use options that are
no longer around or should we maintain behavior as much as possible?
## `CURLOPT_DNS_USE_GLOBAL_CACHE`
This option makes libcurl use a global non-thread-safe cache for DNS if