GSoC 2013

John Marino dragonflybsd at marino.st
Sat Apr 6 01:55:23 PDT 2013


On 4/6/2013 04:43, Mindaugas Rasiukevicius wrote:
>
> Well, do you need an invitation?  I try to answer most of the emails (as
> my free time permits) about NPF.  If DragonFly developers have an interest,
> I am glad to answer questions or suggest on potential difficulties you might
> experience while porting NPF.  After all, this would result in extra testing
> and perhaps contributions to NPF, while DragonFly community would hopefully
> get a decent piece of software.

I've been speaking for myself and not the rest of the DragonFly project. 
  That said, a formal invitation at the beginning of the NPF development 
along with identifying NetBSD-specific interfaces at the beginning and 
implementing abstractions then would have been ideal.  From my vantage 
point, NPF seemed as much "it's for us, we don't care if it's ported or 
not" as PF does.  I'm not accusing you of thinking that way, only that 
this is the final result.

> At this moment, I do not think there are any architectural NetBSD-specific
> parts.  Obviously, NetBSD has different interfaces and you might find that
> adopting some (e.g. ptree(9) interface) might be less problematic than
> converting.  On the other hand, I would consider abstraction or interface
> adjustments to make supporting of a different system easier.

That's kind of you.  I'm sure that both would probably have to occur. 
The DragonFly project have no problem in adapting functionality when it 
has clear advantages.

>
>> <...>  And there's the point that it's not even production ready.
>
> Can you ground this point?  While there are still important milestones to
> complete, I think NPF version which will ship with NetBSD 6.1 has a pretty
> stable feature set.

Mainly I was basing this on the fact that NetBSD 6.1 hasn't been 
released.  People don't use Release Candidates in production.  It's also 
logical to expect that with 6.1 will come a lot of use, and that will 
reveal some issues that the developers didn't discover.  Maybe NPF is 
rock solid out of the gates, and that would be quite an accomplishment, 
but it doesn't have a "CV" yet.

I think you understand that I'm not trying to offend.  It would be great 
for DragonFly if NFP works on it, and I wouldn't be surprised that an 
attempt to make this happen occurs when NPF is proven.  If we had a 
bigger team, then maybe this attempt would have happened by now.

Regards,
John







More information about the Kernel mailing list