[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: strange network-configuration on installation from cd-squeeze-test-amd64-i386-netinst



[Andreas B. Mundt]
> Unfortunately, the network definition in GOsa allows only the
> classical A-, B- and C-networks but not the 255.255.254.0 network
> mask which was used before.

Given that there is no such thing as A B and C class networks the last
20 years or so, that is not really acceptable behaviour for any
software package today.

> (No idea why 255.255.254.0 was chosen in the beginning).

It was choosen to ensure that the skolelinux code network could
co-exist with other networks on the 10.* private segment as long as
they stayed away from the 10.0.2.0/23 subnet.  /23 was selected to get
a reasonable amount of machines (ie around 500).  In hin-sight, we
should have used /22 or /21 to match the larger sites.  The idea when
we selected it was that sites needing larger subnet would have the
skill set to update the network settings, which proved to be partly
incorrect. :)

> So we can either choose 255.255.0.0 or 255.0.0.0 network masks for
> our setup now.  To keep the range of available network addresses as
> flexible as possible, I decided to use a Class-A network mask.  The
> configuration of the tjener interface wasn't modified to reflect
> that, as things are not yet settled.

Using a full /16 network would make it impossible to co-exist with
existing networks using any of the 10.0.* addresses.  But using a full
/8 is worse, and using /24 leave too few addresses in the backbone
network.

> Perhaps someone with deeper insights in the reasons why the network
> was set up the way it is can comment on the issue and help how to
> proceed with classical network masks best.

Fix Gosa to behave like software should have been working the last 20
years is the best option.

Happy hacking,
-- 
Petter Reinholdtsen


Reply to: