[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-net-dev
Subject:    Re: Unix domain sockets (UDS, AF_UNIX) in System.inheritedChannel() and elsewhere in Java
From:       Jaroslav Tulach <jaroslav.tulach () oracle ! com>
Date:       2019-08-30 6:54:57
Message-ID: 4995052.6oXyQqsUfE () dellxps
[Download RAW message or body]

Hello Chris, Alan & all.

Thank you for your discussions with František and thanks for outlining various 
directions to move his inquiry forward. 

> You will need to find a sponsor

Right. I feel motivated to help, as I talked František into contributing 
rather than hacking around the problems he had found in Unix domain socket 
support in the OpenJDK...

> ( a committer on the JDK project )

However I am [just an author](https://openjdk.java.net/census#jtulach), so I 
can't really drive a JEP to create a completely new feature. Moreover I agree 
that ...

> If I'm not mistaken, then this is your first contribution to OpenJDK

... starting contributing to OpenJDK by driving own JEP is a bit too 
challenging. As such I suggest to scale down the contribution as much as 
possible. No new features, no enhancements, just let František fix a bug. I've 
just reported

https://bugs.openjdk.java.net/browse/JDK-8230380

I hope we all agree that letting a Java program read an uninitialized memory 
is a vulnerability and that it deserves to be fixed. I let František come up 
with a webrev and I update the bug accordingly.

Hopefully we then find a committer to apply the fix. Thank you all for guiding 
František through the process of his first OpenJDK contribution!

Jaroslav Tulach
NetBeans Platform Architect
OpenJDK (Graal & Metropolis) Author
OracleLabs

Dne pondělí 19. srpna 2019 12:47:14 CEST, Chris Hegarty napsal(a):
> Franta,
> 
> > On 19 Aug 2019, at 09:34, František Kučera <franta-java@frantovo.cz>
> > wrote:
> > 
> > ...
> > can anyone tell me how to continue? I can change the code (the patch was
> > only proof-of-concept, fast hack) or change the design (e.g. creating a
> > new module instead of patching existing classes) but before I spend more
> > time with this task, I need to know, whether there is interest to have
> > AF_UNIX (AF_LOCAL) support in the OpenJDK (see the first e-mail in this
> > thread for what I mean by "support").
> From my perspective, it would be reasonable to continue to investigate
> the possibility of supporting Unix domain sockets in the JDK. Such a
> feature would likely need a JEP [1].
> 
> If I'm not mistaken, then this is your first contribution to OpenJDK
> ( it's a pretty significant/large one to start with! ). You will need
> to find a sponsor ( a committer on the JDK project ), that is both
> motivated to help you drive the design and implementation, as well as
> having the time available to do so ( I, unfortunately, do not ).
> 
> While this feature does, on the face of it, seem reasonable, in terms of
> priority, it does not bubble up the list of other things that we want to
> do, like; providing support for the newer URI/IRI, improved IPv6-only
> support, RDMA Network Sockets, etc.
> 
> -Chris.
> 
> [1] https://openjdk.java.net/jeps/0




[prev in list] [next in list] [prev in thread] [next in thread]