half_close is pretty obscure

[lvh]

Does it really belong in this PEP?

[jerub]

Lets go with 'no'.

[hynek]

I believe it adds unnecessary complexity/special cases to the initial, supposedly simple PEP.

[lvh]

Okay, taking it out.

[glyph]

I think that you need it, because it's the only way to say "produce some data after EOF", which is critical for a subprocess-based transport.