> Nemo writes from northern Central Park in New York City:
I wish! Still chained to a desk for a now, hopefully that will change.
When I get a laptop, I will be able to write from the Park, weather
-- and muggers -- permitting.
> > Curious about the topology of internetworking in Kenya.... Are the
> > ISPs leasing lines to connect to a single internet hub in Kenya
> > (operated by KPTC?) or are they leasing lines to overseas hubs?
> Better if someone from Kenya answers that, but my impression is that
> they all lease lines through KPTC to overseas hubs. I don't think
> there's a router at KPTC.
> > In the latter case, do they interconnect?
> Ha! The various ISPs in Kenya have a strong reputation for
> collaboration! And it's not a good reputation. But then, isn't that
> what "competitive" markets are all about?
Well competitors can chose to cooperate when it's mutually
beneficial. I would wager that if two significant ISPs in Kenya
inter-connected, their services would be so much faster that the
others would have to follow suit, thus raising the playing field to a
> > That could be a strategy for reducing cost. Two ISPs may lease one
> > long distance higher bandwidth line, and interconnect via a
> > cheaper local leased line. Not only would it be more efficient, it
> > would double the peak speed of the individual user's traffic coming
> > from or going to overseas, and also greatly speed up intra-Kenya
> > communication....
> Agreed that this would be an interesting strategy for coping with
> artificially high government prices, and it would be an interesting
> strategy even if government reduced its prices. But none of the
> firms seem that interested, except for some of the smaller,
> non-TCP/IP service providers. I wonder if such a solution would
> really make that big a difference for intra-Kenya communications, or
> if it would really save enough money to justify the
> transaction/administrative costs of collaboration.
Why should packets going from Ngong Hills to Lavington have to be
routed through Europe or America? If the long distance link is by
satellite, the message will be travelling 36000 x 4 = 144,000
unnecessary kilometers. Even at the speed of light that's half a
second. But even more important than the distance, there are queueing
delays associated with each hop. The far away hub is probably
congested with thousands of other users. The routers are like
airports (without the duty-free shops). There was a time when you had
to transit through Paris or London to fly from one African capital to
another. Not only is it a longer distance, but you have to find your
gate in a huge busy airport, stand in line and check-in one additional
In addition *each* user's peak speed would double or the cost would be
halved: that's multiplexing gain, the main reason for having packet
switched networks in the first place. Given that, unless the total
internet business in Kenya is very very small and will remain so
forever, investing in a single router and inter-connection definitely