Re: 回复： RFR (XS) 8237806: Convert TIMEOUT of sun.net.dns.ResolverConfigurationImpl into property

[Chris Hegarty]

Just to be clear.

The code change that you are proposing affects the JDNI DNS provider, so
may impact DNS lookups performed by JNDI, regardless of the removed-in-9
InetAddress name service provider mechanism. The CSR should be clear
about this.

Given this, the obvious question is whether this now configurable
timeout property should be supported as an environment property, since
that is the more natural way to configure the JNDI environment.

Also, we have not added properties with the prefix `sun.net.`, in quite
a while.

I think there may be a tension here about how best such functionality
could be supported ( if it is to be ), given that the name service
mechanism has been removed since JDK 9. It might be that a potential
JDK 8 solution is not appropriate ( without significant alteration ) for
the mainline development ( currently JDK 15 ), and vice versa.

-Chris.

On 24/01/2020 09:53, Denghui Dong wrote:
> 
> yes, you are right.
> 
> 
> 
> 
> 来自钉钉专属商务邮箱 <(null)>
> 
>     ------------------------------------------------------------------
>     发件人：Alan Bateman<Alan.Bateman at oracle.com>
>     日　期：2020年01月24日 17:34:30
>     收件人：Denghui Dong<denghui.ddh at alibaba-inc.com>;
>     net-dev<net-dev at openjdk.java.net>
>     主　题：Re: RFR (XS) 8237806: Convert TIMEOUT of
>     sun.net.dns.ResolverConfigurationImpl into property
> 
>     On 24/01/2020 09:22, Denghui Dong wrote:
>>     Hi Alan,
>>       In our production environment, the content of /etc/resolv.conf may be modified because of some reason, and we found that JVM can't know it when it running.
>>       There are some ways to fix this problem, one way is Adding -Dsun.net.spi.nameservice.provider.1=dns,sun
>>     to enable sun.net.dns.ResolverConfigurationImpl, but
>>     the TIMEOUT value is too long for us, so we hope to use a system
>>     property to control it.
>>
>     If I read this correctly, you are running with
>     -Dsun.net.spi.nameservice.provider.1=dns,sun so that InetAddress is
>     using the DNS-only name service provider (based on JNDI-DNS), is
>     that correct? Just checking because that mechanism was removed in
>     JDK 9, maybe you are proposing this for jdk8u?
> 
>     -Alan
>