RFC: Too many race conditions.

[Jeroen Frijters]

mark@mark.mielke.cc wrote:
> If .NET 2.0 does do this, then the close() interrupting accept() would
> be broken.

Hmm. accept() appears to be special cased and not use the reference counting mechanism. Instead they check if the handle is closed after accept() returns and then close the handle returned by accept(). This prevents it from being a security hole, although there is small probablity for a DoS.

Regards,
Jeroen