std::io::net may need more thought

[alexcrichton]

The net module has a few things which I would like to consider improving:

• Must the Acceptor and Listener traits exists? It's a shame having to import the traits just to make a simple server.
• Does the layout make sense? std::io::net::ip is quite long. Possibly a top-level net module? Possibly shorten using reexports?
• What more needs to be exported? The current primitives implement the basic Reader/Writer traits, but not much beyond that. There are many methods provided by librustuv/libnative which are not exported. Need to make sure the signatures are correct.
• "Unix pipes" are not unix pipes on windows (they are named pipes)

Wish list

• TcpStream::open("google.com:80") does not work
• I can clone a tcp stream, but I cannot close that duplicate tcp stream (from my owned stream)
• Creating a server is quite wordy. There are a number of imports, lots of ip addr configuration, lots of listening/accepting, etc.

Nominating, I believe it is quite important to have a solid networking story.

[sinistersnare]

Does net need to be in std?

I guess this is me agreeing that there should be a top level net module...

[liigo]

And std::io::fs too long too, std::fs? libfs? After std facade…

[sfackler]

TcpStream::open("google.com:80") would be pretty straghtforward to implement, but we'd need to pull liburl into libstd most likely.

[alexcrichton]

I believe that with the "libstd facade" that net does belong in std, but it perhaps belongs in its own crate behind std (such as a libio and a libnet), but these are just mild musings of me.

I do think that a "libstd facade" will make implementing TcpStream::open("google.com:80") nicer to stomach.

[pnkfelix]

Marking P-high, not 1.0 milestone. Specific bugs can/should be opened if there are particular items that you want to assign higher priority (i.e. want to get them on the 1.0 milestone).