RFR 15 8243099: SO_INCOMING_NAPI_ID support

[Alan Bateman]

On 01/05/2020 14:19, Chris Hegarty wrote:
> :
>
> Has any consideration been given to restricting the set of allowable
> integer values that may be set for this option - say, to none? Which
> would then lead to an IllegalArgumentException if `setOption` is
> invoked in an attempt to set this socket option - since no given value
> is allowable.
>
This is a good topic and I think you are right that this is the first 
time that an option will be specified as "read-only". There are corner 
cases with a few of the existing socket options are awkward. SO_FLOW_SLA 
comes to mind as one that doesn't have a default value so can't be read 
until it is set. The SO_BROADCAST system option can't be enabled on some 
operating systems without special operating system permissions.

I think the choices for the exception are UOE or IOException. I don't 
see IAE as an option because it implies that there is at least one value 
that is valid for the parameter and this isn't the case here. If UOE is 
used then it would require tweaking the javadoc a bit to allow it be 
thrown when the option is not writable. IOException might be the least 
worst option as it's already thrown in several cases where socket 
options can't be changed (for various reasons).

-Alan.