Re: Regression in svnserve: IPv6 causes IPv4 not to work [was: Branching for 1.3.]

Nicolás Lichtmaier wrote:
>
>> My attached patch differs from his that users must explicitely
>> enable IPv6 if they want svnserve to listen on an IPv6 socket. It passes
>> the svncheck and svncheck6 (using --ipv6) on OpenBSD/i386 -current.
>>
>> Existing IPv6 users of svnserve have to add --ipv6 as option to
>> svnserve, so for them this patch is a regression.
>
>
>
> IMO this is not correct. The ipv6 API is designed so that its use is
> automatic, and people get IPv6 without even knowing about it. Yes, there
> have been issues now and then, and some programs included a "ipv4-only"
> switch. But the default should always be the future proof solution. I
> then propose a "--ipv4-only" switch that could be used when people with
> weird configurations have trouble...

FWIW, my experience with CUPS is that it is possible to default to
IPv6 for typical configurations (listening on all interfaces) and
disable the IPv6-only mode to allow IPv4 clients to connect as well,
transparently. The only time this doesn't work is if the user has
specifically disabled the IPv4-to-IPv6 gateway stuff in the kernel.

My vote (not that I have one) would be to use the IPV6_V6ONLY socket
option when it is defined, and to add a --ipv4-only switch as you
have proposed. That will allow for non-conforming configurations to
continue working in IPv4-only mode while providing out-of-the-box
IPv6 support.

-- 
______________________________________________________________________
Michael Sweet, Easy Software Products           mike at easysw dot com
Internet Printing and Document Software          http://www.easysw.com
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org