Inccorect use of rfc2616 in basic_headers about beast HOT 6 CLOSED

This brings up an interesting question, what should the behavior of basic_headers::replace do? There are a few choices:

1. Replace the first occurrence
2. Replace the first occurrence, remove any additional occurrences
3. Replace the first occurrence if there is only one occurrence, else throw an exception.

from beast.

Another problem, what happens with basic_headers::operator[] if there is more than one field with the same name? I guess it could return the first occurrence.

from beast.

i would remove operator[] and instead do something like equal_range in multimap or maybe return a vector with all values.

i would say replace should do remove(name) insert(name, value)

from beast.

The problem with removing operator[] is that it inconveniences the majority of use-cases where only one field with a given name exists. For example:
https://github.com/vinniefalco/Beast/blob/master/include/beast/websocket/impl/stream.ipp#L970

Also note that the fields in the headers are in insertion order and not in set order. This is an important detail, because proxies are not supposed to reorder header fields. Many other HTTP libraries make the mistake of returning the fields in alphabetical order, which breaks proxies.

from beast.

This seems to pass tests, it changes basic_headers to behave like a std::multiset
https://github.com/vinniefalco/Beast/commits/rfc

from beast.

Fixed in 1.0.0-b8

from beast.