adding rsockets support into JDK

[Lu, Yingqi]

Hi Alan,

On the JEP, I will merge your feedback and Chris's feedback into the document today. Since this is my first time drafting a JEP, what is the next step I need to work on besides the webrev?

On the patch, I will follow the example of DatagramChannelImpl connect/accept for RDMA based channels and modify the code in the next revision.  

Thank you!

Lucy

>-----Original Message-----
>From: Alan Bateman [mailto:Alan.Bateman at oracle.com]
>Sent: Monday, May 21, 2018 3:45 AM
>To: Lu, Yingqi <yingqi.lu at intel.com>
>Cc: nio-dev at openjdk.java.net; Viswanathan, Sandhya
><sandhya.viswanathan at intel.com>; Aundhe, Shirish
><shirish.aundhe at intel.com>; Kaczmarek, Eric <eric.kaczmarek at intel.com>
>Subject: Re: adding rsockets support into JDK
>
>On 18/05/2018 17:06, Lu, Yingqi wrote:
>> Thank you for your feedback, Alan!
>>
>> I will modify accordingly, update the JEP online and send a copy to jdk-dev
>for broader review.
>>
>I read through the updated JEP (in JDK-8203434) and it reads well now.
>
>One suggestion is to split the Description so that the API is in its own section
>after the Description, and before Testing. That would give you a place to
>explain that the feature is platform specific / non-standard so the proposal is
>to add the API to the jdk.net module which is where JDK-specific networking
>features are exposed. The text you have after "Below is the list of proposed
>public APIs" can follow.
>
>One other thing is Risk/Assumption #2 where -
>Djava.net.preferIPv4Stack=true is needed. We updated the implementation
>in JDK 7 to allow for mixing of IPv4 and IPv6 sockets in the same VM and I think
>you should be able to make use of that and avoid needing to force all sockets
>be IPv4 sockets. The example to look at is DatagramChannelImpl as you can
>create a DatagramChannel to an IPv4-only socket. In the RDMA
>SocketChannel/ServerSocketChannel implementations then you should be
>able to specify the protocol family to connect/accept. You don't need to do
>this in a first version of course but I think addressing this point will make it a lot
>more usable.
>
>-Alan.