Hello Lynwood;
As the author of such I'll try to answer your concerns.
On Sat, 2016-08-20 at 19:39 -0400, lleachii--- via 44Net wrote:
I agree about the security concerns, etc. But in some
uses of IP, it
would be very difficult to provide services to multiple devices over NAT
technology (e.g. allowing multiple servers needing the same TCP or UDP
port or IP protocol number, providing an adjacent private network the
option to route to the host via direct connection or over the Global
Internet). Also, the reason supporting the statement is quite
inaccurate. A large majority of ISPs not yet implementing Carrier Grade
NAT, do in fact, assign the end user pubically routable IPv4 space.
Granted, it's commonly a single /32 IP address. The page actually goes
on to describe the need of IPs according to the same principal - which
is basically today's standard practice.
You're reading too deeply into for whom the page was written for.
There's no mention at all about security configurations nor NAT. It's
drafted for the most novice of users, not those who have a good
understanding about the protocols we use. I think while you were reading
it, you were injecting your own interpretations of things that don't
exist in the page.
The comment regarding 802.11 devices (or any IP
routing device for that
case) is somewhat dubious security-wise ("this would not include any
802.11 routers for use on /HamWan/HamNet/ as doing so would make you
quite insecure."). Archives of security comments in this forum from
others suggest proper firewalling is necessary in environments running
IPENCAP-enabled routers, ESPECIALLY BECAUSE of the presence of
NAT/masquerade co-existing in some AMPRNet nodes. It can only be assumed
from the Wiki that the Ham allocated the IP space ensures or guarantees
firewalling by using NAT. While NAT has security benefits, it was
invented with an affect of slowing IPv4 exhaustion. In our IPENCAP
environment, it is a vector to send IP datagrams to "this network" (i.e.
how your router perceives 0.0.0.0/0), AMPRGW, other gateways, or an
adjacent private network (e.g. your home/corporate/government LAN). For
this reason, the OpenWRT router setup page includes information
regarding configuration of a basic Virtual Routing and Forwarding
environment on your Linux-based router. This configuration places any
AMPR interface on a separate, routing table not possessing locations of
non-AMPR subnets (i.e. routing table "44"). While I'm not certain if any
node in the /HamWan/HamNet/ networks run IPENCAP, others in AMPRNet do.
I think you're injecting terms that don't exist on the page and are
trying to squeeze them in where a novice would be greatly confused. As
you know, once you decode the incoming IPEncap frames into your lan, you
no longer need to continue routing those frames via IPEncap and you can
easily push routes via RFC-1918 space internally within your lan. Also
one who's running BGP would have a totally different configuration than
that you suggest and would not be running IPEncap. While ISPs may give
end users a single /32 IPv4 IP, their backbones aren't necessarily
public IPs. With most of the ISPs I've been with, we've used 10-net
space with the routers we interconnect within our intranet network with
various border routers having public IPs to our peers. If this didn't
work, we'd have had an excessive amount of very upset customers and if
we assigned publically routable IPs on our backbone routers we'd be
exposing ourselves to those we wish not to try and crack into. Again,
this would be way too confusing for the novice with no IP experience and
most likely would turn them away from Amprnet due to their lack of
knowledge.
It's important to note, without a firewall, and
even with a Virtual
Routing and forwarding Instance, these inbound packets would have
forwarded to AMPRGW. Unless an operator arranges to use private
addresses directly with me, these packets were invalid because my tunl0
interface faces the Global Internet - where use of these IPs are not
allowed by RFC 1918. Accordingly, I possess no specific route on my
"table 44" that AMPRNet nodes should use to reach these destinations.
Other tools to improve security are programs to automate adding and
flushing entries on a firewall permitting AMPRNet endpoints to send
IPENCAP. Even then, a mis configured client or infected machine within
AMPRNet could send a crafted packet to traverse our network, or networks
physically or logically adjacent to our nodes.
As you know, a route table alone when you're hosting multiple route
tables does not tell a frame how to take it's path. It's the policy
that's in place (via ip rules) that tells the frame which route table to
use. A firewall doesn't set policy routing, just frame filtering. If
you're using RFC-1918 space within say an 802.11 network the edge (as
with ISPs) would still have to maintain some sort of a public IP for
global routing purposes whether it's for IPEncap or BGP, and RIP would
have to announce the path.
This of course is not something you would need to configure, nor would a
novice. This would be on the one who's engineering such a network
configuration of 802.11 nodes. Someone using RFC-1918 space would have
to somehow announce their network within the RIP system so the path to
them would be known to you using a globally facing IP as their "gateway"
into that subnet. For example, if I had 44.250.250.0/24 via 1.2.3.4, I
could route the whole /24 on a network of 802.11 routers using RFC-1918
space and whatever 1.2.3.4 is for a device would have to either be a BGP
facing gateway router OR could be an IPEncapsulation system that can
chop up my amprnet /24 subnet based on specific RFC-1918 space hosts.
You do nothing except allow RIP into your daemon for your policy routing
as you do now. As a matter of fact, I am doing that now with multiple
802.11 routers that use 192-net RFC-1918 space and it works just fine.
44.88.0.0/27 routes internally to 192.168.1.1 which then splits that
block into two /28 subnets and pushes one of them to 192.168.2.1. You do
nothing different on your end and you can see the entire block.
As you mention about firewalling, security is the responsibility of the
specific individual that's hosting a subnet of 44/8, which includes
anti-spam rules, port filtering, etc. It appears a lot of linux folk
like using fail2ban. If you're not encapsulating on your lan and assign
an M$ machine a 44-net IP then again it's the individual's
responsibility to insure the security of that device. (Ok so that's an
oxymoron considering the golden key crack was released)
What I may suggest is if you wish to do a write up on the issues that
you're concerned about, make the page and we can link it to the
"Requesting A Block" page. There's so many different ways to do security
it'd be a good challenge (ie: fail2ban, gShield, hardware devices, etc)
--
I was going to go see a Guns 'n' Roses concert...
until the liberals took away ALL the guns. Now I'm
only going to see Roses *sigh*
-----
73 de Brian - N1URO
email: (see above)
Web:
http://www.n1uro.net/
Ampr1:
http://n1uro.ampr.org/
Ampr2:
http://nos.n1uro.ampr.org
Linux Amateur Radio Services
axMail-Fax & URONode
http://uronode.sourceforge.net
http://axmail.sourceforge.net
AmprNet coordinator for:
Connecticut, Delaware, Maine,
Maryland, Massachusetts,
New Hampshire, New Jersey, Pennsylvania,
Rhode Island, and Vermont.