On Mon, Jul 30, 2007 at 07:52:47PM +0200, Tomas Pospisek wrote:
> * Software that binds to the first socket found
> Then there's software that binds to the first port it gets and is
> difficult to teach not to do so. [2]
There is nothing like a "first socket". Software either binds to any
address or to a specific address. Some software may lookup the "first"
ip, this is rather broken.
> * The cost of disabling IPv6
> Once the kernel has loaded the ipv6 module, one can not get it rmmod'ed
> (or not easily - I have not figured out how to do this remotely on a
> hosted server). Which means:
It is not possible.
> * Limited usefulnes of IPv6
> I would guess that less than one in a thousand users have direct access
> to an IPv6 network. Getting connectivity to IPv6 is still non-trivial
> (based on my own personal experience).
Less than one in thousand servers uses Fiberchannel, should we drop
support for them?
> * ask the user at install time whether he wants IPv6 on
> * disable IPv6 by default and make it easy to re-enable
> * make IPv6 *easy* to disable
Some arches already have ipv6 compiled in.
Bastian
--
One does not thank logic.
-- Sarek, "Journey to Babel", stardate 3842.4
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
