David Buchmann

i agree on the direction of your draft, yes. i was thinking of refactoring the cache control into its own class out of HeaderBag because HeaderBag is growing too big....

that topic would be covered by the RFC with targeting: https://www.rfc-editor.org/rfc/rfc9213.html#name-cache-behavior it would need some design decisions how to do that in the application however, as it would not be...

hm, good question indeed. it looks like one can only add additional directives to respect, but not remove no-cache and max-age. this had been added like this with the original...

oh i see. you could extend the (Builder)[https://github.com/KnpLabs/php-github-api/blob/9afaf87f99c6c2acde80b6925d1b97a70ece8cb5/lib/Github/HttpClient/Builder.php] and overwrite the addCache method. while i agree that the current behaviour is unexpected and strange, i am afraid to change it...

unfortunate timing btw, we just tagged 2.0 2 weeks ago. otherwise we could have declared it one of the important BC breaks of 2.0. we could add a new factory...

we could switch this client to be a PSR-18 client. but that would mean somebody would need to commit to maintain this client. there are a bunch of http clients...

if we decide to keep this maintained, we should adjust the socket-client.rst page in the documentation, reverting https://github.com/php-http/documentation/pull/310#discussion_r1193398598 and maybe add more information.

update: if we go PSR-18, we can use httplug/discovery which since recently automatically installs a PSR-18 compatible client if none is specified by the application.

closing in favor of #443

good idea. it should be part of a "migrating to httplug guide"