ROFL, zing!
Telkom also uses SAFE as a backup, so what is he talking about?
When Telkom had an absolute monopoly, and could charge what they liked for as little bandwidth as they liked, they only sold "restorable" international circuits, using SAT-3 or SAFE as the primary route, and failing over to the other cable automatically in the case of a failure. This has a number of unintended consequences.
Firstly, about 70% of South Africa's traffic is to Europe and North America, and only about 30% to India, the Far East and Australia. When SAT-3 fails, the 70% of traffic has to go via SAFE, and back to Europe via SEA-ME-WE3, adding nearly 150ms of delay in the process. That's why any ISP using Telkom's traditional wholesale circuits suffered from poor performance. (As reported, this automatic restoration failed as well, basically cutting some ISPs off altogether.)
Secondly, this cable restoration is an expensive business, since the entire backup capacity has to be purchased up front, not just on SAFE, but on SEA-ME-WE3 as well, which is a notoriously expensive (and long) cable system. In Telkom's defence, that's one of the reasons that their services used to be so expensive (quite apart from their money-grubbing monopoly habits). It was all bundled together, so you couldn't tell what you were actually paying for.
Restorable circuits via long routes are all very well if all you are running is phone calls and private data circuits (like Telkom used to). Today, the majority of traffic is IP (the Internet), and routes, traffic flows and latency matter a lot. Hence, the global trend is towards the use of non-restorable circuits, carrying point-to-point IP, with no backup. (Even Telkom is now offering these, in response to competition in the market.) The world's four or five Tier 1 ISPs build vast meshes of these point-to-point links between and across continents, and, being IP, any one link fails, the next packet just takes the next shortest path, with no noticeable impact for most users.
Hence, SEACOM has changed the paradigm in South Africa. With three physical routes heading in three directions, we now have the start of a proper mesh to make us part of the real, global Internet - it's ultimately all a question of the speed of light, and the shortest great circle distance around the globe. We don't (yet) have direct cables to South America or Australia (or Antarctica, if there's anyone who cares), but such cables would improve matters even further. SEACOM is the shortest route to East Africa, India and the Middle East, but is also almost the same length as SAT-3 to Europe (look at a globe). Hence, for that 70% of traffic, it's a far better alternative route than SAFE.
Of course, the worst thing a local ISP can do is to buy a single, non-restorable, point-to-point international leased circuit to some faraway place to connect to Tier 1 Internet. There are still South Africans (including some ISPs who should know better) who think that the Internet is in London or New York
and not a global mesh network, but this increasingly untrue. It's very widely distributed across all continents. If all an ISP has is a single route like this, not only will they get poor latencies to much of the world, but they also run the risk of being off the air for weeks (the repair time of a deep sea cable fault on any one cable).
Hence, an ISP can either build their own global Tier 1 network (a somewhat costly exercise, probably in the region of several $bn), or they can connect to an ISP that has one, with a node in South Africa. Obviously, this probably doesn't mean all that much to users on any of the 3G/wireless networks, since most of their latency and performance is in the air interface, but it certainly matters to local ISPs, and to corporate customers, particularly those who make use of servers outside South Africa.
