The Myth of "Dialtone"
By davew on Dec 04, 2006
Way back when - it must have been about 2001 - Scott McNealy bought into the concept of "webtone", where the idea was that "IP network access must be as reliable as the dialtone on your conventional telephone".
The thing is, the Internet may well already be more reliable than that.
It's a subtle point that the limited format of telephone tones means that they can be used to hide shortcomings in the 'phone network.
For instance, if you pick your 'phone up and get a dialtone, all that means is that you have connectivity to your nearest PBX. In IP terms, this is exactly equivalent to being able to ping your default gateway (ie, is only one step better than being able to ping your own address).
If you dial a number which is also on your local PBX, then it's about as easy for the PBX to make a circuit connection to the other handset as it is for a router to route packets between directly-connected subnets; the only usual difference is that the PBX needs to watch for a hang-up signal so that it can drop the connection, but it should also be borne in mind that such active communication termination makes billing easier. Even then, there are subtleties involved; it used to be the case that CCITT and Bell differed where one would only drop the circuit when the calling party sent a hangup, whereas the other would only drop when the called party sent a hangup.
If you call a number outside the scope of your local PBX, then the usual thing happens where the PBX reaches up to its nearest Class 5 switch - at which point, the traffic "disappears" into the cloud of interconnected Class 5 switches in the same way as an IP request disappears into the cloud of interconnected Internet routers when you try to connect to a remote IP address.
However, if the appropriate Class 5 switch can't set up a call circuit as a result of link congestion, you get an engaged tone.
That's right, an "engaged" tone. Not a "switch unavailable" tone or a "link full" tone, nor do you have to sit waiting interminably as you would if you were on a browser in the IP world, while whatever "connecting..." telltale spins away to itself for a couple of minutes before giving you something informative along the lines of "connection timed out" or "server unavailable".
If you've ever tried 'phoning someone, had the line come back as engaged, called them again some minutes later and had them react with surprise and denial when you ask who they were talking to the first time you called, the link was actually saturated the first time.
This also varies between countries; when making international calls, sometimes I've even had a "ringing" tone to mask conditions of line saturation, rather than an "engaged" tone.
The only time you would typically get a "number unreachable" tone is if the Class 5 switch which provides the nearest link to the destination-number holding PBX is unreachable on its out-of-band signalling link, and as Class 5 switches are deployed in highly-available pairs (just like resilient IP network infrastructures), this doesn't happen very often.
The fact that connection failure reports in IP networks are so informative is what makes failures more apparent, and hence perceived reliability lower.
Mobile 'phones are at least starting to change the rules a bit; when a hex (or BSC, if you're in the game) reaches capacity, at least you now get a "Network busy" error on your 'phone's screen rather than just an engaged tone or a dropped connection attempt. Folk will begin to notice such bandwidth issues, and therefore dynamically-modifiable BSCs with adaptive coverage are being developed (Queen Mary College, University of London have some particularly nifty stuff).