The article in your February issue on how to use a Japanese keyboard in English Windows 95 is very useful to those who are using Japanese notebooks. On the other hand, using the Registry Editor is not for every one.

-- Paul Spee


Interesting column on NTT's OCN service [March, page 16]. I have mixed feelings, however. As an OCN customer myself, the more press I see about the service, the more I worry -- because I know for a fact that the rumors of massive overselling are true. I am in the process of turning up my fourth OCN Economy 128K connection, and have been using the service for about 6 months.

During that time, I have spoken with several people involved in the operation of NTT's OCN division network [and] your estimates of how much they are overselling bandwidth are optimistic. However, they are constantly deploying new bandwidth... and researching ways to distribute bandwidth more effectively (queuing and port level traffic shaping).

Although this means connectivity through OCN Economy service isn't a death spiral, it does mean that as they try to catch up with sales, transit load will fluctuate. [I have seen] peak time loads that caused a 3-second latency through their network. Holy saturated routers, Batman! However, I have faith that while they are obviously shooting for profit-driven growth, they do intend to scale.

I need to take issue with two points made in the column, however. Mr. Romero states that, "Based on observable packet loss, it seems likely that the OCN network relies heavily on frame relay." Any IP network faced with more ingress than transit bandwidth will drop packets.... Frame relay technology is no more the cause of IP packet loss than any other encapsulation protocol; even the much-touted ATM discards cells when a circuit becomes congested. Therefore, while your deduction is correct, your analysis methodology is wrong.

My other point of contention is that Mr. Romero slags on OCN's customer support personnel. While it is true that they are not allowed to discuss much of the engineering that OCN Economy is using, this is not unique to NTT. Try calling up Sprint and asking what they are using to hub their internal ATM mesh, or what release of IOS they are running.

My personal experience with NTT's OCN customer service representatives over the last six months has been consistently exemplary. I have always seen a prompt response to my correspondence -- in English. Perhaps they felt more inclined to help since I was an actual customer.

Overall, I have been extremely satisfied with my OCN Economy service. It is an excellent value and, aside from congestion, highly reliable.

-- Matt Ghali

The author responds:
Thanks for your comments. The paragraph you quote should have read: "observable latency" rather than "observable packet loss"; the error was mine. I have, however, since confirmed that OCN does indeed make extensive use of frame-relay.

My point about NTT's refusal to answer questions about total available bandwidth and the number of users sharing it was that this information is critical in evaluating the quality, and potential future quality, of such a service. I recommend caution when dealing with any provider that refuses to furnish this information.

-- Tim Romero


Back to the table of contents


Send your comments, criticisms, and kudos to editors@cjmag.co.jp