Category Archives: Uncategorized

ethernet over 3G/GPRS

As weird as it sounds, it is exactly that, ethernet bridge over 3G/GPRS. If you want to make your own, you will need: two machines running Linux, each with one ethernet interface and one 3G/GPRS interface, OpenVPN and a little of configuration.
First, create tap interfaces on each end (check OpenVPN docs on how to do that)…
Then bridge those with ethernet interfaces…
Now bring up your 3G (in fact, it is even easier if you have fixed connection on one end as OpenVPN server needs fixed IP)…
Next, configure OpenVPN over your tap interfaces without IP addresses and/or routing…
Start OpenVPN and see how everything coming in on ethernet interface of one of your linux boxes is going out on another and vice versa.
As for me, I used two Teltonika RUT-104 devices, slightly improved OpenWRT and SIM with fixed IP on one end. Of course, if you can not get SIM with static public IP, you can use DDNS and dynamic IP, but it still has to be public and of course, as mentioned above, server end of business does not have to be on 3G at all, it can be anywhere on internets and even behind NAT, as long as you can forward OpenVPN traffic to it.

Asterisk 1.6.x app_fax

It is broken a little bit. Well, not exactly app_fax, but as it is the only foss component using chan_sip/udptl one might conclude that app_fax is the broken one. In fact it is chan_sip T.38 negotiation that is broken. It does not take in account that udptl does not have one very important variable initialized and relies on optional attribute in sdp to set that variable. If this attribute is not sent by far end, udptl spits out warning and goes on allocating buffer of size -1. Which leads to segfault mentioned in my previous post.

Stick with what works

So you have made a decision to go on with some particular system for some particular task? Good. As long as it works. But if it doesn’t for one reason or another, dump it. Use investment principle – if it is at loss of 5% (or whatever margin you have set) don’t expect it to come back and make you profits, just sell it and look for another, better option. The same applies to software, if you are struggling with the system and it calls for more and more siliconpower and does not give you returns you were expecting, don’t wait for miracle to happen, just look for another software package/consultant that is more likely to deliver what you need and write off whatever you spent on the first, non-working system.