On Sunday, 19 February 2017 at 04:00:33 UTC, Gavin wrote:
On Friday, 6 May 2016 at 08:32:03 UTC, Robert burner Schadek
wrote:
As discussed yesterday at DConf, curl in phobos must go.
The plan is as follows.
1. undocument everything curl related in may 2016
2. deprecate everything curl related in may 2017
3. delete everything curl related in may 2018
3.1 move curl stuff to undead
PR: https://github.com/dlang/phobos/pull/4283
Two of the most important modules in a standard library are
containers & networking. In Phobos they are terrible. How about
focusing on that than this nonsense? The Curl wrapper is about
the only useful bit of networking code in the entire standard
library. How on earth are you gonna compete with Go & Rust with
useless container & networking modules?
I'm of the opinion that we should grow the curl wrapper.
Take what is done with python for example: their most famous
library ever is requests, and anybody doing networking in python
first reaches for it. What is the standard library urllib module
for then? Well it's the basis requests is written on, it's meant
to provide basic, low-level access to the network, and that's ok.
People just learned to use requests.
Right now unfortunately the curl wrapper isn't a very good
building block to build libraries but by growing it into
something with better ranges etc it could be. The critics of the
API and memory usage don't need to be: just grow transparently a
second facade to the library. No code broken, better tools for
better libraries and happier users.
The only real point left is managing the C code in phobos but as
Andrei said (not in this terms) if Phobos can't do it what good
is the "interfacing with C is dead easy" pitch?