Is `keep_alive` intentionally not implemented for `header`?

[inetic]

The keep_alive function is implemented for message but it only requires information from the header. Due to this, any function that composes a response header needs to take the whole message unnecessarily.

Our work around is to just re-implement the keep_alive function and use that, but I'm wondering if there is a reason I'm missing for the exclusion or whether is just hasn't been a priority to have it in header.

[vinniefalco]

The implementation is actually in header. I don't know why it is in message. Try patching Beast and see if it works? Maybe we could merge it.

[ashtum]

I think the reason might be that request_header and response_header are two separate classes, and keep_alive is common to both request and response messages. Moving it to the header would lead to code duplication. As for why it is not in basic_fields, it might be to keep basic_fields limited to functioning as a container.

[vinniefalco]

the implementation of keep_alive() depends on knowing if the message is a request or a response, which is not possible in basic_fields.

[ashtum]

the implementation of keep_alive() depends on knowing if the message is a request or a response, which is not possible in basic_fields.

I think keep_alive is a common field in both request and response and the implementation doesn't check for that either. However, it requires the message version, which is only available in request_header and response_header:

beast/include/boost/beast/http/impl/fields.hpp

Line 817 in 9b24b28

get_keep_alive_impl(unsigned version) const