Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 23 Jul 1999 19:01:41 +0900
From:      itojun@iijlab.net
To:        Jeroen Ruigrok/Asmodai <asmodai@wxs.nl>
Cc:        hackers@FreeBSD.ORG
Subject:   Re: Will FreeBSD ever see native IPv6 ?? 
Message-ID:  <9128.932724101@coconut.itojun.org>
In-Reply-To: asmodai's message of Fri, 23 Jul 1999 10:39:42 %2B0200. <19990723103942.C9276@daemon.ninth-circle.org> 

next in thread | previous in thread | raw e-mail | index | archive | help

>> 	We (KAME) are using 3.2-RELEASE and 2.2.8-RELEASE because we can't
>> 	base our IPv6 development on top of moving target.
>> 	FreeBSD 3.x-STABLE and 4.x are moving target (which moves very quickly)
>> 	and are unusable as base version for us - if we need to chase two
>> 	moving things (IPv6 and FreeBSD) we are doomed.
>Problem being is that the 2.2.x branch has been declared obsolete and dead
>by the Project.
>If serious development work is being done CURRENT has always been the
>targetted platform. But let's not nitpick about decissions make months
>earlier.
>However, 3.2-STABLE isn't that fast moving forward, and I doubt, based on
>earlier discussions about touching the 3.x branch, that Jordan will allow
>IPv6 to be entered into the 3.x branch (however TPTB may judge otherwise).
>So 4.x seems like the only alternative left...

	I can't tell KAME users like: "Oh, if you would like to apply KAME
	diffs you'll need to fetch FreeBSD 4.x of June 1st".  KAME needs to
	stick to x.y-RELEASE either, KAME have no choice anyways.

>And I am still working on that =)

	Thanks for the effort!

>> 	There has been NRL/INRIA/KAME integration work going on (basically
>> 	to avoid "4 BSDs and 3 IPv6 = 12 choices" nightmare by making one
>> 	IPv6 stack).  There are, mainly, some (or too many) management
>> 	issues there.  We will be resolving management issues issue very soon,
>> 	hopefully by next week.
>That's good new ITOJUN-san, but I hope you can understand that after seeing
>OpenBSD deploy IPv6 and NetBSD-CURRENT getting the IPv6 code merged by you,
>that FreeBSD (read the userbase/developers) is anxious to also incorporate
>this.

	I talked with core@freebsd.org several times about IPv6/IPsec
	integration, and the reaction was that:
		FreeBSD can wait till unified-ipv6 is made available, since:
		- IPv6 is not that urgent task and
		- it will be messy if FreeBSD integrates KAME first,
		  then switch to unified-ipv6.
		  (making a big change twice for IPv6 is questionable route)
	So we are in the current situation.  (or maybe I was not pushing hard
	enough)

	NetBSD merged in KAME code last month, because NetBSD will have feature
	freeze for 1.5 soon and needed IPv6 code earlier than that.
	If I miss 1.5, it will not be able to be merged until 1.6 (planned to
	be sometime late 2000, I heard).  NetBSD will be switching to
	unified codebase whenever it is made available (so NetBSD decided to
	make a big change twice).

>> 	There's incomplete "unified" codebase there, which is not very
>> 	ready for public consumption.  Anyway please hold till the managment
>> 	issue is resolved, I believe I can give you a good news.
>Let me ask one question: if a few of us got the 3.x kit of kame up to 4.x
>level and it got all commited would it make the `merged' stack easier to
>integrate into CURRENT or does the stack-project prefer to use a clean
>codebase? I myself think the first, which allows our driver writers time
>to adjust to IPv6 changes, get a lot of still-present bugs out and basically
>restabilize the system before the next IPv6 changes. Not to mention allow
>people to test/work with IPv6 already.

	Here's my question: how much patch rejection you saw when
	you tried to apply KAME/FreeBSD32 patch onto 4.0?  Were there
	many of these, or very few?

	If they are very few, and core@freebsd is okay, I think KAME
	can be merged into 4.0 tree first, then FreeBSD can switch 
	from KAME to unified whenever it becomes ready (just like NetBSD does).
	Hopefully KAME and unified-ipv6 will not be that different,
	except for dual-stack tcp part (KAME code used separate tcp4/tcp6
	for a long time, due to maintenance and stability reasons.  The way
	unified code does dual-stack tcp is different from what KAME does).
	Userland part can be reused without trouble.

	There are, however, several drawbacks in doing that:
	- Now we have multiple KAME/FreeBSD[34] repository, one is in
	  KAME site and one is in freefall.freebsd.org.  Synchronizing
	  those tree is a big problem.
	  KAME already have problems synchronizing code between platforms
	  we support, the merge will increase the problem.
	- Manpower issues.  We need to continue providing KAME/FreeBSD32
	  anyways, for people who wants more stable IPv6/IPsec systems.
	  We can't just tell everybody to switch to 4.0 and chase
	  FreeBSD-current.
	- We need some more FreeBSD commit privs for other KAME guys.
	  (this is a easy part)
	- again, two big changes for IPv6?  Impacts on IPv4 stability?

itojun


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9128.932724101>