←back to thread

79 points bertman | 1 comments | | HN request time: 0.303s | source
Show context
buttocks ◴[] No.45138851[source]
In case you think fax is now ancient history, there are tens of millions of faxes still crossing the PSTN every day. Fax is alive and well!
replies(2): >>45139664 #>>45141471 #
ajb ◴[] No.45141471[source]
They don't work the same way they used to though. Since old school PSTN hardware is being replaced with VoIP, there is a hacky protocol called T.38 which does just enough to convince each side that it's talking to a real fax, and decodes and forwards the data over IP.
replies(2): >>45142022 #>>45142576 #
psim1 ◴[] No.45142576[source]
T.38 is actually a fine way of transporting fax bits but unfortunately it is quite uncommon to see T.38 end-to-end. While a VoIP provider may negotiate T.38 with a customer's fax ATA, it is likely being transcoded to G.711 by a gateway at some point as it traverses the telephone network, ultimately making T.38 a less-reliable choice. (Better to have the same codec end-to-end.) A comparison might be cellular or VoIP providers offering wideband codecs, which sound great when you stay on-network, but when the call crosses the PSTN and is transcoded, the sound is worse than if you used the standard narrowband end-to-end.
replies(1): >>45143350 #
1. ajb ◴[] No.45143350[source]
There's nothing wrong with the wire protocol, but T.38 was designed to work with faxes that don't know it exists. Given that VoIP doesn't provide the same highly-constant latency as the PSTN, T.38 gateways pull some dodgy stuff, such as (if I recall correctly - I no longer have access to this code) deliberately introducing HDLC CRC errors to give themselves time to wait for a packet to appear.