Status of the RDMA code

[Alan Bateman]

On 29/02/2020 10:48, Chris Hegarty wrote:
> Thank you for sending this update Ivanov.
>
> Looks like very good progress, and nice to see the PR’s have been 
> accepted and merged into rdma-core.
>
I agree, it's good that these issues have been fixed in rdma-core.

Overall I think the API is in good shape and the RdmaSockets class 
description is clear on the SelectorProvider methods that it implements 
and the limitations around interop.

One aspect of the implementation that could be re-visited is the socket 
adaptors and the SocketImpl. The socket adaptors for 
SocketChannel/ServerSocketChannel have changed significantly since Lucy 
started the rsocket effort so one thing that could be done is to refresh 
the equivalent in the RDMA code and then see if they could be used as 
the RDMA Socket/ServerSocket. It would mean the SocketImpl in the webrev 
could go away. There is a small point of differences related to how 
socket vs. channels work with Thread.interrupt but that isn't hard to 
deal with.

Stepping back for a moment, this might be a good time to re-assess 
whether this feature needs to be in the JDK. Given the spec work to 
separate it from the default provider then it should be possible to 
separate its implementation from what of java.base, meaning that it 
doesn't rely on access to sun.nio.ch classes or any other JDK internals. 
I think it would be a good exercise to try this and see what issues come up.

-Alan