> >/The list of UDP ports at Wikipedia(1) doesn't show port 93 or 10093, /> >/so I suggest we correct Wikipedia's list, since port 93 is shown as /> >/assigned to "Device Control Protocol" by the IANA(3). /
> Yes, our usage of port 93/UDP for AXUDP is completely unrecognized by
> IANA. There is no official document describing AXUDP so there is no
> way for us to claim a port - although there's no prohibition on
> using any port number we want, since port numbers are not absolutes.
> Still, it's not wise to double-up on port numbers.
In the "old days" it was quite easy to register port numbers, and in fact
I have registered a few myself that I still use for similar but not the same
purposes (459, 1535, 1536). However, it appears to be not as simple anymore
and looking at protocols that are in much more widespread use than AXUDP it
appears to be not common anymore to register ports used only by some proprietary
protocol at IANA (there are way, way more protocols in use than those with
IANA-registered port numbers).
Note that at the time I registered the above 3, IANA would not take suggestions
for the numbers to register. I claimed different numbers that were unregistered,
but they gave me those instead. Asking for port 93 to be registered is unlikely
to succeed, 10093 may be possible.
Registered port numbers are useful mainly for open protocols that are widely
implemented and that accept unsolicited incoming connections. So everyone knows
what port you have your wellknown services listening on.
There is not much that can go wrong when using a port number without registering
it, especially when the receiving application accepts data only from a
pre-configured peer. The port numbers used essentially don't matter in that
case, the only thing you need to make sure is that you don't have another
session with that same peer that would want to use the same port number.
(e.g. it would not be wise to use port 5198/5199 as these are used by Echolink,
another port usage that is not registered at IANA and is in much more widespread
use than AXUDP)
Rob
> I'm a bit confused by your post: since 172.29.x.x addresses are
> reserved for "Detached" networks, please clarify the source of the
> 172.29.60.0 IP range you mentioned. Is it provided by your ISP, or are
> you going through a corporate network to get to the Internet? TIA.
This range is from RFC1918 where the 192.168 addresses also come from.
He probably has those addresses on his LAN and uses NAT to connect to internet.
Of course it is possible to use NAT between such a range and the AMPRnet, but
I don't recommend it. It is better to assign net-44 addresses to AMPRnet devices,
maybe have 2 addresses on one device.
Understanding a mixed-network NAT configuration (NAT to internet and NAT to AMPRnet
in the same network router) is hard for most people, and often mistakes are made,
resulting it lots of RFC1918 traffic ingress in the AMPRnet.
I get lots of log entries like this in our gateway:
Mar 3 09:56:13 Packet DROP: IN=gre8 OUT=tun0 SRC=192.168.1.101 DST=44.137.25.148 LEN=590 TOS=0x00 PREC=0x00 TTL=60 ID=0 DF PROTO=UDP SPT=5060 DPT=5060 LEN=570
(you see, the timestamp is "current")
Apparently someone experimenting with SIP. My own SIP phone is at 44.137.41.104
so I don't have such issues, I can connect to that 44.137.25.148 SIP server without NAT.
Rob
I've bitten the bullet and started to make some progress getting on 44net.
I'm certain my current issues are firewall and I'll get it sorted out. My
non-ham network is in the 172.29.60.0/22 range. I don't think I'm leaking
internal stuff to my partially built tunnel but If anyone sees packets from
that block please let me know.
Status: Attempted to ping 44.0.0.1 on a system on the ampr enclave on my
network, With tcpdump listening on the gateway ampr0 interface I can see
packets going out on the interface but that doesn't mean the packets are
getting out of the box. I'm not seeing a response but like I said, I think
it's a firewall issue.
I just wanted to say Hi to to everyone and say I'm looking forward to
exchanging packets with everyone
73 de n2xu
Tom Cardinal/MSgt USAF (Ret)/BSCS/Security+ ce
--
73 de N2XU/Tom Cardinal/MSgt USAF (Ret)/BSCS/Security+/IPv6 Certified
> Unfortunately, there are problems with AXIP and AXUDP ever since the
> revisions of AX.25 were published - the AX.25 frames may get too big
> to encapsulate in unfragmented IP and UDP frames. Any new document
> would have to discuss the fragmentation issues, and I'll bet they're
> not defined. (As far as I know, the same issue applies to ROSE, KISS,
> and Net/Rom encapsulation too.)
Is this an issue? The AXUDP protocol simply is "put AX.25 frames in UDP
datagrams" and an UDP datagram can be nearly 64KB in size so I would not
think there are any issues. UDP and IP handle the fragmentation and
reassembly and the receiver gets the original datagram to insert it into
the AX.25 stack.
Of course, AXUDP senders should not set the DF flag, and if they do, they
should not expect more than about 512 bytes maximum frame size. Which
still is enough for classic AX.25 packet radio.
> Do any of the various implementations of AX.25 encompass v2.0 et seq as
> published by TAPR? Ie, is anyone actually implementing jumbo frames,
> expanded frame sequence numbers, etc? Or should we just treat it as a
> bad idea that should be buried and go on with the old protocol?
The actual AX.25 version and extensions like jumbo frames and extended
sequence numbers should not affect AXUDP at all. The UDP layer only replaces
the HDLC layer normally used to transfer data over radio, and all things
happening in AX.25 should be compatible between the endpoints, as always.
Rob
