Closed Nyholm closed 4 years ago
FYI there is a similar and related conversation in the oauth2-client
Yes, thank you for the link.
What do you think of this PR @stevenmaguire?
Hi @Nyholm,
Firstly, thank you very much for this PR! Wow. I love it.
I'm going to look further into the code to see if I can provide any feedback and I'll checkout the discussion that @stevenmaguire referred to.
Watch this space :)
Thanks for the PR @Nyholm :+1: for me
I think the implementation is fine from a "how" perspective. I am not completely sold on the justification for the "why" perspective. I've been following the conversation in the oauth2-client project, as well as glimpses of similar conversations in other projects where HTTPlug is being heavily marketed, and I am not sure that the "problem" being solved here is actually a problem worth solving. With respect to the conversation in oauth2-client, I tend to agree with the points that @shadowhand has made.
My vote, if it matters, is no.
Hi @Nyholm, so obviously this discussion is way stale. But I have an update.
In the new version, we have completely decoupled from any particular HTTP implementation. We're using PSR-7, PSR-17 and PSR-18 only. We recommend the use of Guzzle, but we don't require it.
Thank you for opening this discussion years ago, it's definitely been factored in with the rewrite.
To completely decouple from Guzzle I've used the HTTPlug abstraction. This change does not break BC.
The benefits of this is users stuck with Guzzle5 can still use this library.
FYI: other libraries using HTTPlug are Geocoder, Mailgun, FosCache, KnpGithubApi and more.