STATE OF THE JAPANESE INTERNET, 1997:
SUCH DISTANCE TRAVELED, SO FAR TO GO

by Bruce M. Hahne
Tokyo (February 20, 1997)

I spent a day at the "Net & Com '97" exposition at Makuhari Messe last week, trying to soak up some useful technical information admidst a sea of booth staffers waving marketing survey clipboards in my face. The show, sponsored by Nikkei Business Publications, used to be called "Open Systems Expo / Network Expo", but has now been renamed and given an internet / intranet spin in order, presumably, to attract a wider base of exhibitors and attendees. Whether this strategy was successful or not remains an open question, since my rough estimate of the day's traffic through the convention floors put the ratio of booth staff to conference attendees at roughly 1:1... certainly a ratio bound to warm your heart if you're an attendee looking for personal attention from a salesperson, but less so if you're an exhibitor looking for a good marketing return on your booth expenses.

As this was the last network show that I expect to attend in Japan (I'm presently laying plans for a return to the U.S.), and as I've been known to occasionally inflict my opinions onto an unsuspecting public, I thought I might provide this essay as a parting gift, or at least a parting stirring-up of the waters, to the electronic community which keeps an eye on network developments, and particularly Internet development, in Japan. Perhaps by looking at where we've been, and where we are, we might gain some insight into where we'd like to be in the future and how to get there.

** Too many choices **

Unlike the U.S. Internet, where engineers are presumably regularly heard complaining loudly when their ping times between Boston and San Jose push above 20 milliseconds, in Japan you can count yourself lucky if your round-trip ping time to California is under 300 ms., not including the heavy additional delays involved in going through your modem. The basic leased line speed in Japan, after all, is still the 64Kbps link, 1/24 the speed of the "plain vanilla" 1.5Mbps T1 link so common for leased-line Internet access in the U.S. Some Japanese providers have a T1 to their upstream site these days, but the majority are still working off of links somewhere in the 64Kbps to 512Kbps range. You'll get 40 ms. ping times just going across a single 64Kbps link. Still, even though it seems sometimes that the Internet world (or at least the traffic) in Japan moves at 1/24 speed, there's still a demand for bandwidth. If anything, the bandwidth problem is worse in Japan, since the equipment available to the end user is the same speed available anywhere else in the world (28.8Kbps analog, 64Kbps for ISDN), but the domestic IP backbone isn't as built up as it is in the west.

One issue that Internet providers are going to have to deal with is the basic question of which speeds to offer. Back when the 28.8Kbps "V.FAST" pseudo-standard came out, the joke was that the creators had nicknamed it "V.LAST", since that was the last amount of bandwidth anyone would be able to squeeze out of a POTS (analog) line. That got bumped up to 33.6Kbps not too long afterwards, and now it's about to go to 56Kbps downstream, 33.6Kbps upstream. In the meantime, we still have the older 38.4Kbps "baby ISDN" equipment, i.e. just about every external ISDN terminal adapter on the market. This gear looks to your computer's external serial port like a standard asynchronous modem (and in fact it usually does a reasonable imitation of the Hayes AT command set), but it connects to an ISDN line and expects a similar piece of ISDN TA equipment to answer its call at the ISP side. This type of connection was popular back before the wide availability of true 64Kbps ISDN cards. The baby ISDN equipment has now been geared up to 57.6Kbps, still running out your PC's serial port, still compatible with essentially none of the standard central-site equipment that ISPs like to use. Nowadays, we're seeing more of a push towards true 64Kbps ISDN, and once users get the hang of that they're going to start kicking and screaming for 128Kbps bonded ISDN, which means that we will then have reached the point where one dial-up Internet user will be able to single-handedly swamp the upstream leased line connection of a large percentage of the ISPs in Japan.

Now, counting using my fingers and toes, I see 28.8Kbps (analog), 38.4Kbps (ISDN), 56Kbps (analog), 57.6Kbps (ISDN again), 64Kbps (ISDN, one channel), and 128Kbps (ISDN, 2 channels bonded) as the full set of speeds which a truly dedicated and overworked ISP engineering department in Japan might choose to offer to its end users. Coming up with a pricing plan for just one of these speeds is a pain in the neck, and billing for it is a hassle, particularly if you choose to meter instead of charging a flat-rate price. But setting price points and creating a metered charging and billing system for SIX different connection speeds is enough to make any ISP throw in the towel. Matters are made worse by the fact that RADIUS, the authentication, access, and billing protocol used by any ISP worth its salt, and the most likely candiate protocol for bailing ISPs out of the multiple-speed mess, has only recently made it beyond the "Internet draft" standard in the IETF, plus Ascend has hacked up the draft standard with its own attributes to such a degree that we may never see convergence between Ascend and the rest of the RADIUS-speaking world. And anyway, typically the ISP doesn't get enough information back from its central-site chassis to be able to determine, after the fact, whether a user's connection was at 28.8Kbps, 64Kbps, 33.6Kbps, 57Kbps, or 1200 baud for that matter.

Without a doubt, the answer is simplification, and my prediction is that we will see pricing for all speeds from 28.8Kbps to 64Kbps merged into a single entry on the typical ISP's pricing sheet, with 128Kbps commanding a premium. There will be some holdout ISPs, typically those which don't have PRI-based central-site equipment, which try to provide a 3-level pricing system consisting of 28.8Kbps, 64Kbps, and maybe 128Kbps, but when your competitor down the street is offering 56Kbps service for the same price as 28.8Kbps, it's going to be hard to maintain the distinction.

** Speed: too much is never enough **

Japan is well positioned to be able to offer 64Kbps and 128Kbps connections to the end user due to the reasonably easy availability of ISDN connections in major cities. You still have to pay the usual highway robbery rate of Y72,000 plus construction charges per line (for the yen-illiterate: current exchange rates are about Y125 per US$1.00, so after construction charges we're talking about $700 per individual phone line) that you install into your home or business, just like you do for an analog line, but at least if you're in Tokyo, Osaka, or somewhere else large, you have a good chance of getting an ISDN line within 3 weeks after you order it. However, to go beyond 128Kbps speeds to the end user, you have to start looking at other technologies. The recent APRICOT (Asia Pacific Regional Internet Conference on Operational Technologies) conference which I had the good fortune to attend in Hong Kong in January presented a series of sessions spelling out what the main options are: XDSL, Internet via cable TV feed, and Internet via satellite. XDSL is high-speed transmission over copper phone lines, and requires equipment located at the phone company's central office to tap into the copper. This technology is getting a lot of attention in the states these days, but if it's ever offered in Japan, and I haven't heard any rumors that it will be soon, you can expect it to only be offered by NTT, since they control all the copper. Internet over cable TV lines is seeing some dabbling in Japan, but due to the low penetration of cable TV into Japanese households and the typical need for the cable company to replace existing "downstream only" central-site equipment with new two-way equipment, I don't expect to see easy availability of Internet-over-cable services this decade either. Internet over satellite means that the consumer actually buys a dish, points it towards the sky in the right direction, plugs the cable from the dish into an interface card on his/her PC, and fires up the web browser. Typically, the satellite feed is an incoming-only feed TO the consumer, with an analog phone line used simultaneously as the backchannel connection allowing the end user to transmit Internet packets. This technology presents some difficulties in Japan since transmission to satellites is, like every other form of communication technology in Japan, heavily (over)regulated. "However," I was recently speculating to myself, "what if some company were to move the satellite transmission offshore, broadcasting from Australia or California and simply selling the end-user dish equipment in Japan?" And, in fact, this is exactly what we're going to see within the next 6 months. Press releases have announced that Direct Internet, a joint venture including Hitachi Cable, Sony Music Entertainment Inc., and Japan Telecom, will be providing high-speed (300 to 400 Kbps) Internet connections to end users in the Asia-Pacific region, including Japan, via a satellite uplink from Napa, California.

** Telco bashing **

Bandwidth within Japan, and from Japan to elsewhere, remains a pricy problem. Domestically, NTT pretty much has a lock on local loop circuits, plus they own almost all of the relevant national infrastructure, plus the government doesn't seem particularly intent on forcing some pricing sanity into the monopoly areas of the market. Since I've always wanted to copy out some of NTT's local loop pricing for posterity, hoping that perhaps some day in the distant future some net archeologist will stumble upon my writings and have a good laugh at what things must have been like in the bad old days, I'll take the opportunity to do so here. To avoid copying NTT's entire tariff chart and putting everybody to sleep, I'll stick with 64Kbps, 256Kbps, and T1 pricing. These are the recurring prices; installation and construction charges aren't included, nor are the prices of renting NTT DSU's at each end.

----------------The Only Table In This Essay------------

NTT leased line prices for various distances.
Prices are in thousands of yen per month.

to 15kmto 30 kmto 50 kmto 100 kmto 160 kmto 200 km
64Kbps*53**104132140147150
256Kbps#137##275376410433447
T133765577795610801160

*Y53,000/month at present, but will increase to Y77,000/month over the next 15 months.
**Y104,000/month at present, but will increase to Y113,000/month over the next 15 months.
#Y137,000/month at present, but will increase to Y200,000/month over the next 15 months.
##Y275,000/month at present, but will increase to Y307,000/month over the next 15 months.
--------------------------------------------

The first thing you're likely to notice about these prices is that they're high. The average Internet provider in the U.S. is paying perhaps $500 per month for a T1 local loop from a baby bell to connect to its upstream backbone provider. In Japan, a 25km T1 local loop costs Y655,000 plus DSU rental charges per month, putting us close to $6000/month for that T1 circuit. Basically, U.S. ISPs, when I hear you complain about how much you're paying your telco for your local loop, I just don't have any sympathy. We're paying 12 times that rate here in Japan.

The second thing you might notice about the NTT leased line tariff chart, if you had the entire sheet in front of you, is that the speeds on the chart max out at 6 Mbps. There IS no pricing on NTT's standard leased line rate chart for the T3 (45Mbps) lines which the U.S. Internet backbone has criss-crossing the country, and let's not even think about higher speeds like OC3. Once you go beyond 6 Mbps (and, in reality, it's actually once you go beyond 1.5Mbps), you're off the chart, lost in the NTT bandwidth twilight zone of "maybe it's available, maybe it's not, and if it's not available today then you can just sit on your hands until we're good and ready to upgrade our equipment to serve your needs". Given this situation, it's no wonder that many of the more well-funded Internet providers in Japan have chosen to locate their equipment inside the KDD building, where they can get high-speed connections to KDD's international transmission equipment without having to purchase corresponding high-speed NTT local loops.

The third thing you might notice, if I had put the 192Kbps speeds into my chart, is that there's a large price jump, inconsistent with the 64Kbps pricing, between 128Kbps and 192Kbps. This is because for leased line speeds above 128Kbps, and for ISDN PRI connections, NTT uses fiber exclusively, all the way to the equipment in your office. Never mind that installing fiber typically requires you or NTT to rip up the street, punch a hole in the wall of your building, run new conduits through your ceiling, and drill holes in the interior building wall to mount a special NTT fiber containment box; never mind that getting that fiber from NTT to you will take a bare minimum of 3 months (4 to 6 months is more common, and that's in Tokyo) the first time you pull it in; never mind that T1 speeds and ISDN PRI could easily be run over 4-wire copper without requiring any of the above delays or expensive construction. NTT has decreed that 192Kbps and up shall be run over fiber, and so fiber it is. The result is that when a Japanese Internet provider outgrows its office space and its bandwidth and needs to look for a new location, it gets to play the "find the bandwidth" game. This game typically involves calling as many NTT managers as you know and trying to divine from them which of Tokyo's 23 wards actually has the fiber AND the NTT central-site switching capacity to handle your immediate and future needs. Once you've chosen your ward, don't forget to ask NTT to bless your building as well... if you choose a building far away from the fiber, they may have to tear up the street, and it takes a while to get a permit to do that (regulations, you know). If you're particularly unlucky, there might be a strip of "national roadway" (just about any numbered highway will do... there are several within Tokyo) between your building and NTT's closest fiber drop. This situation will require an additional round of internal NTT and governmental paperwork while NTT obtains the right to do construction to pull your fiber under the national roadway.

Over the few years that I've been in the Internet business in Japan, I've seen all of the following behaviors from NTT:

- Insufficient planning has caused chronic shortages in many locations. Back in 1994, it took 3 months after an order was placed to receive a 64Kbps leased line from NTT, in central Tokyo. We're thankfully over that problem in Tokyo now; usually you can have your 64Kbps line in under 4 weeks these days. But Shinjuku-ku remains a truly lousy location for someone trying to pull high-speed lines into your building. Yokohama is rumored to have similar problems.

- NTT will never, ever, commit to an install date for a leased circuit or a PRI circuit. If you're lucky, you'll be able to pry a good guess out of them for a 64Kbps leased line, and a target month out of them for an INS1500 line. For the higher-speed lines (anything running over fiber), NTT will as likely as not call you a few days before the line is projected to go in and tell you they've postponed it by another month. At the first Internet provider I worked for, we got the "we've postponed installation again" treatment for several months from NTT before we finally got all of the INS1500 lines we had requested before we even moved into the building. The victims of the delays, of course, were the end users who were getting busy signals every night due to our phone line capacity shortage.

- NTT will look at you with a straight face and tell you that it will take 9 months or more to get basic INS64 (copper-based ISDN) into many semi-remote areas. Since they're the only ones who can provide it, there's nothing you can do about it. There's nobody else you can take your business to. Sure, Japan has ISDN... if you live in Tokyo, Osaka, maybe Yokohama. If you live in Kurashiki, Omiya, or anyplace else that NTT hasn't bothered to install sufficient INS64-capable switching equipment, be prepared to wait. A long time.

- NTT will do what it can to prevent its competitors from reaching you. I suppose this is more of a regulatory problem than an NTT problem, since it's in the nature of all money-loving businesses to try to squash their competition. Back in 1995, when we were looking at pulling our first fiber into our building, one company we talked to was TTnet, a struggling competitor to NTT in the Tokyo local loop market. TTnet did some checking and determined that in order to run their fiber into our building, they would need to string it across the phone pole that sat right outside our window, a few meters from the building. It turns out that in Japan, that phone pole was not considered a public right-of-way point; it was NTT property, and NTT denied TTnet's request to run fiber over the pole. Sorry, came the reply back to us from TTnet, but we will be unable to provide you with service... NTT won't let us connect to you.

Ah, you say, but the recent restructuring of NTT should solve all of these problems. At last, Japan will have real competition in the marketplace. Actually the much-ballyhooed "breakup" plans which hit the press in December are, if anything, going to strengthen NTT's power while not doing anything for the consumer. The typical press blurb from December '96 on this topicran something like this: "NTT will be divided into two domestic companies, with a third company providing international service. All three companies will be held by a single holding company. The restructuring plan is subject to government approval." Here, for the benefit of the reader who doesn't speak NTT-ese, is what this press announcement actually means:

- "NTT will be divided into two domestic companies"
Translation: we will redraw a few of the lines on our management chart to present the impression to the outsider that there are two distinct operating companies.

- "with a third company providing international service."
Translation: we also get to enter an area of business that we were forbidden to compete in before. Now we can exploit our domestic monopoly to provide international end-to-end solutions that undercut the pricing structures of other international carriers. All we have to do is keep our interconnect prices to competitors high, just like we've already done successfully to domestic competitors like DDI and TWJ.

- "All three companies will be held by a single holding company."
Translation: this isn't really a breakup. There will be centralized management giving all of the orders and making sure that all three companies work together.

- "The restructuring plan is subject to government approval."
Translation: the government will pass laws WEAKENING Japan's anti-trust legislation to allow us to create the holding company.

In the words of GLOCOM researcher and long-time NTT watcher Stephen Anderson, "The debate is over and NTT has won." I'm afraid I have similar sympathies. The next time somebody tells you that NTT has been "broken up", you have my permission to laugh.

** International restrictions **

For international leased lines connecting to Japan, only type 1 carriers can have facilities-based services. There are only three type 1 carriers (KDD, IDC, ITJ), and their international half-circuit prices are regulated by the government. Today, the Japanese side of a U.S.-to-Japan international leased circuit will cost you about twice as much as the U.S. side of the circuit. Of course, most Internet providers aren't even in a situation where they have to worry about half-circuit prices, since with the standard "type 2" telecommunications license held by the vast majority of ISPs in Japan, purchasing your own international leased circuit, or IPL (international private line), is illegal. In order to buy an IPL from one of the type 1 carriers, you have to have a "special type 2" license. In order to receive a special type 2 license, you have to have money (a few hundred million yen would be a good start) and know which bureaucratic strings to pull with MPT. Since most ISPs don't have anything close to hundreds of millions of yen in working capital, typically the best they can hope to do is buy from a special type 2 upstream such as IIJ or Tokyo Internet. Such a heavily controlled environment contrasts starkly with the situation in most other nations. In Hong Kong for example, any ISP is allowed to lease its own international circuit to anywhere, including to the U.S. Internet backbone. Some have done so, and some haven't, giving the Hong Kong consumer a wider range of options when choosing an ISP.

** Wanted: more exchange points **

One way to reduce the amount of IP traffic that you, as an ISP, send internationally is to connect to an Internet exchange, or IX. At an IX, multiple providers connect to a common high-speed LAN and route packets to each other, reducing the traffic on more expensive international or long-haul leased lines. Japan presently only has two IXes, both located in Tokyo, with a third in the planning stages in Osaka. All of them are run under the guidance of WIDE, Japan's academic Internet system. NSPIXP1 is the original IX and allows connection rates of up to T1. NSPIXP2 is modelled after the higher-capacity IXes in the U.S. and allows connection speeds of 45Mbps. Not surprisingly, NSPIXP2 is also colocated inside of the KDD building, allowing providers who already have equipment located there to connect to NSPIXP2 without paying NTT for a 45Mbps local loop.

For at least a time, IXes in Japan were at best in a legal grey area, since Japan places heavy restrictions on interconnecting networks. (Never mind that interconnecting networks is what the Internet is all about...) Participation in NSPIXP1 was carefully phrased as "a collaborative research project" sponsored by WIDE, and the recurring connection charges paid to WIDE by the providers at the IX were "joint research fees". Such phrasing apparently allowed the IX to pass beneath government radar. Given the fairly obvious non-research nature of NSPIXP2, it would seem that IXes are something the government is willing to accept, but it is telling that there are still only 2 operational IXes in Japan today, both operating under the umbrella of WIDE. Boardwatch magazine counts 13 IXes operating today in the U.S., and they've missed some of the newer regional IXes such as the Atlanta Internet Exchange. Perhaps Japan, with its "Tokyo is the center of the universe" mentality, will never need more than one IX in Tokyo and one in Osaka, but it would be nice to see a few more.

** OCN terrors **

I could scarcely claim to be writing an essay about the Internet in Japan circa 1997 without a mention of NTT's "Open Computing Network" (OCN) plan for getting into the Internet business. Some have gone on record as saying that OCN will have more of an impact on Japan's Internet in 1997 than any other driving force. I tend to disagree. Rather, I'd say it's likely that the FEAR of OCN will have more of an impact than any other driving force. In the past 6 months, almost certainly due to fears that OCN will slash and burn leased line Internet prices, we've seen two of the formerly high-and-mighty backbone leased line providers do an about-face and suddenly announce that they're now serious about competing in the dial-up PPP business. AT&T Jens is rolling out the AT&T WorldNet dial-up service in Japan, and in the best "how low will the mighty stoop" case study I've seen in a long time, old-timer IIJ has been taking out full-size billboard and magazine ads for its new "IIJ-4-U" dialup service featuring two nude women (don't worry, Senator Exon, they've cropped the photo before we get too far down below the neck) and one computer. This from IIJ, "the backbone provider with an attitude and a mile-long investor list", as I like to call them. So what's the deal, IIJ? Do I get the nude women when I sign up for your dial-up Internet service? Or have all your marketing people just been reading too many issues of Young Jump?

To me the whole OCN fiasco looks like just another face on the same old telco monopoly game: charge your competitors more to access your network and compete with you than the price that you're charging directly to the consumer. Although OCN involves both NTT's entry into the dial-up and leased line markets, the dial-up pricing is harmless; 15 hours per month for a few thousand yen per month, with 9 yen per minute after you go over your hourly limit during the month. This pricing is similar to that of hundreds of other Japanese ISPs. It's the dedicated Internet connection pricing that has the community up in arms: Y37,000/month for a 128Kbps Internet connection, INCLUDING the leased line; Y350,000/month for a T1, and Y980,000/month for speed freaks who want that right-side-of-the-tariff-chart 6 Mbps connection. Now, if you check NTT's standard (non-OCN) leased line prices, you'll find that a vanilla 128Kbps end-to-end connection of under 15 km will cost you Y74,000/month plus DSU rental charges. This Y74,000 is the same price that your friendly neighborhood ISP has to pay to NTT if you purchase a 128Kbps link TO the ISP... and that doesn't include the Internet port charge that the ISP has to tack on to cover its own Internet bandwidth costs, staffing costs, equipment, and profit margin. What we're seeing in OCN's pricing, then, is confirmation of something I've believed for a long time: NTT's leased line pricing is too high. It simply makes no rational sense to charge Y74,000 per month for a point-to-point 128Kbps leased line, while charging half of that price for the leased line PLUS the value-added service of Internet connectivity, a service which ostensibly requires heavy international bandwidth, a robust domestic network, a Cisco router port at the NTT central site, heavy investment in Cisco 7000-series equipment to act as default-free backbone routers, and a team of trained (and extremely rare) Japanese Internet router gurus to manage the whole thing.

A dedicated OCN connection, by the way, grants you a maximum of only 5 IP addresses, which is enough to connect 4 computers and one router. You can supposedly have 10 IP addresses if you beg. This is enough for the small (very small) office to connect, but if you have more than 9 computers that you want to put onto the Internet, you're out of luck with OCN. I've also heard this nasty rumor that OCN will be run entirely over zero CIR frame relay... and believe me, as someone with extensive experience running Internet packets over zero CIR frame relay (I have since learned the error of my ways), if this is what OCN is doing, you may want to look elsewhere. Zero CIR means that the entire network has no guaranteed bandwidth.

On the other hand, every time I hear some more juicy details about OCN, the details have changed, or the rollout has been postponed, or it's all tentative, so by the time you read this they may have changed the playing field again. Regardless of what happens, it's put the fear of God into many of the providers which have a heavy base of leased line customers (IIJ and Tokyo Internet come to mind). The situation reminds me of not so long ago when Tokyo Internet hit the scene in April 1995 with Internet dedicated circuit pricing that undercut IIJ's pricing by a factor of two. Back in those days, if memory serves, IIJ was asking Y400,000/month for a 64Kbps port on its routers, and Tokyo Internet came out with a 64Kbps port price of Y198,000. As the months crept by, every other provider had to reduce prices to stay in line with Tokyo Internet's pricing... even mighty IIJ had to announce some reductions. Eventually, Tokyo Internet dropped the pricing again, to an unheard-of Y98,000/month for a 64Kbps port. Today, a 64Kbps Internet port in Japan will set you back between Y60,000 and Y200,000 per month, depending on who your upstream provider is. It's true testimony to the old mantra that competition works. As someone who used to sit on the board of directors of an Internet provider, I must admit that I saw a bit of red every time I heard that Tokyo Internet was adjusting its prices downward again. I'd wander around muttering vague threats about what I'd do to Toru Takahashi, head of Tokyo Internet, if I were ever to meet him in person. However, quite honestly, Toru Takahashi and his price wars have probably done more for the accessibility of Internet service in Japan than any individual since Jun Murai.

** New toys **

The Net & Com show, while not as large as Networld/Interop Tokyo, the traditional "event of the year" to attend for those in the Japanese Internet industry, still packed an impressive lineup of big-name companies and consumed a fair amount of floor space in Makuhari. Sun, HP, NEC, Microsoft, Lotus, and Novell all took out large chunks of floorspace, as did a variety of NTT spawn including NTT Data, NTT International, and NTT PC.

When I'm at shows, I'm a network hardware person, not a software person. If what you have on display doesn't have a back panel that I can plug cables into, I'm likely to quickly move on to the next booth, and after I've seen the fifth or sixth vendor with an enterprise-ready client-server software product available today for Windows NT or the Unix server of your choice, my eyes start to glaze over and I start looking desperately for something that I can hook up to an ISDN line. Thankfully, even though the newly-named Net & Com show isn't truly an Internet show, or even a networking equipment show, there were enough toys on display to keep me happy. One of my favorites, it being the first time I had seen it, was the US Robotics Edge Server card (more appropriate would be "server on a card") for the USR Total Control network chassis. Sure, we all know you can fit 4 modems on a card that slides into your rack-mountable box... but this double-width card is a full Windows NT server-on-two-cards, including a floppy drive, 800 MB hard drive, VGA port, keyboard port, serial port, 64 Mb of RAM, and a 100 Mhz DX4 Intel CPU all in one slide-it-in-and-it-works package. The idea behind this is "why put your web server in a big external box on your ethernet when you can put it INSIDE your access server?" Now, if only if it were running some OS other than Windows NT... I didn't ask the booth staff if I could throw out NT and install BSDI Unix instead.

** Standards wars **

And while I'm on the subject of modem vendors, I noticed that the 56Kbps-over-analog wars are starting to heat up in Japan just as they are in the U.S., with USR pushing its X2 technology, and some literature from Rockwell conspicuously nearby at a different vendor's booth pushing Rockwell's "K56Plus" technology. Both do the same thing for you, of course: 56Kbps downstream to your modem so that your web pages come in faster, same old 33.6Kbps upstream to your Internet provider, with two catches: first, your provider has to be using central-site equipment which taps directly into something digital, meaning that in Japan your provider will need to be connected to NTT via ISDN, usually PRI (NTT calls this "INS1500"). Second, both you and your provider have to use compatible equipment for you to be able to get 56Kbps service. USR X2 won't work with Rockwell K56Plus, and vice-versa. Frankly, in Japan this is a battle that I expect Rockwell to win, at least in the ISP market, because the overwhelming majority of ISP central-site equipment that uses INS1500 today is Ascend Maxes, and Ascend uses Rockwell chips. To use USR modems at 56Kbps, your Internet provider has to be using USR Total Control equipment, and by USR's own count there are only a handful of providers in Japan which have deployed the USR TC. Advice to USR: if you want to sell the Japanese public on 56Kbps modems for personal Internet use, you're going to have to crack the ISP market. To do that, you're going to have to offer an aggressive ISP discount plan to give providers a reason to defect from Ascend, which has a 2-year head start on you.

** Toys and trends **

The show also gave some hints of future directions for Japan's Internet. First, the general trend seems to be that Internet-related products in the west are making it to Japan faster than they used to. I was a surprised to see that Internet-to-TV technology, which was making its mark in the U.S. this past Christmas in the form of the "WebTV" product, was on display at Net & Com and is apparently available to buy today. JCC demonstrated the "super iBOX" Internet TV appliance (basically a stripped-down PC with an NTSC video connector), with list pricing that starts at Y54,000. Of course, they were cheating by running the video connection out the S-VHS jack into high-end Sony wide-screen televisions, giving a much higher video quality than the average user would see on an average TV set, but the fact remains that the technology is here and ready to go.

Security products, mostly firewall hardware and software, were on display at a large number of booths, and I saw displays advertising encryption cards at several locations. If this is any anticipation of demand, we may see a lot of sites throw up firewalls in the next 12 months. Whether they'll actually be configured to offer any reasonable security is another issue, but at least the purchasers will FEEL safe.

Also present were the usual large number of low-end "SOHO" ISDN and leased-line routers. What's interesting at this particular point in time is that we're seeing so many vendors with these basic 64Kbps and 128Kbps routers that they're starting to seriously think about price as a competition point. For at least a year, Yamaha has dominated Japan's 64Kbps/128Kbps router market, because they were the first vendor to get out a product which had all of these features:

1. It worked as advertised.
2. It had documentation in Japanese.
3. It handled leased line connections as well as ISDN connections. 4. The price was right; \200,000 list initially, now down to about \139,000 list, which means you can buy it for about \100,000.

Ascend, which had an early lead in getting these routers to market in Japan, fell behind in my opinion primarily because they delayed providing Japanese-language manuals. Tzone in Akihabara, for example, sells Yamaha routers, not Ascend routers. New on the scene, however, is the Cisco 760 low-end router, and Cisco is rather surprisingly adopting an aggressive pricing strategy. I've seen this product, which doesn't handle leased line connections yet as far as I can tell, priced LOWER than Yamaha's rather low pricing. Is the Cisco brand name enough to overcome Yamaha's momentum? Time will tell.

The big surprise of the Net & Com show was the large number of voice-over-IP products on display, suggesting that at least the technology, if not the regulatory environment or the business will, to shift voice phone calls onto Internet connections is becoming available in Japan. Granted, the fact that Net & Com specifically wooed Computer Telephony equipment vendors for a special section of floor space managed to draw some of these products out of the woodwork, but still... voice over IP? Only 12 months ago that idea was considered a fringe hobbyist experiment even in the states, and now we're seeing IP-capable PBXes on display at a business technology show in Japan? The mind boggles. Just to reinforce my state of mind, hiding within one of the demo booth sections I found a one-page flier from Rimnet which says that they're using technology from a company called Vienna Systems to provide true PSTN-to-PSTN connectivity running over the Internet. I was informed two days later that Rimnet has started offering voice telephone service betweenTokyo and Osaka, presumably running the voice traffic as IP packets over its own leased circuits, for rates which undercut those of NTT... and then informed two days after THAT that MPT has just declared Rimnet's PSTN-to-PSTN voice IP service illegal. Politics as usual, it would seem... if it's good for the consumer but NTT doesn't like it, it's not going to happen. It's interesting to compare MPT's "no Internet telephony" attitude to the situation in Australia, where there is an active Internet telephony service run by OZemail providing real competition to the established telcos. I know it's providing real competition because I heard a Telstra manager complaining about it at APRICOT in Hong Kong two weeks ago. What Japan needs is its own equivalent of the VON (voice-over-network) coalition (www.von.com) to push in favor of MPT rulings allowing full interconnection between Internet networks and the public voice network. It might teach NTT a thing or two about its own long distance charges.

** My advice **

Nobody ever takes my advice, but I'll give it anyway, if only so that I can say "I told you so" a few years from now.

To small Japanese ISPs: Some observers believe that you're doomed; that OCN, declining margins for dial-up, and the advent of advertising-sponsored free Internet services will put you all out of business. I'm not convinced, nor is Jack Rickard of Boardwatch Magazine, who wrote in September that "customer service and scalability are the only issues that matter in providing Internet access. The big telcos and cable companies are not going to 'take over' internet access and drive the little guys out any time soon." Providing the hand-holding necessary to get the newbies up and running with their first Internet connection is tough, time-consuming, and not something that the big boys necessarily want to deal with... after all, there are an awful lot of newbies out there, and sometimes they can be a real pain to teach. If you can maintain a strong reputation for personal service in your local area, and keep your overhead costs down, I think there's a future for you. However, no matter who you are, your customers WILL start asking for 56Kbps and 64Kbps dial-up services. This means that if you haven't already done so, you need to study up on ISDN-capable central-site equipment and start thinking about ordering an INS1500 line from NTT. If you don't operate in a major metropolitain area, order that first INS1500 line NOW... you'll force NTT to haul the fiber to your building, and by the time they get it there you'll probably be wanting that line.

To Yamaha: your basic 64/128Kbps ISDN and leased line RT100i router is nice, and your four-line central-site version is nice too, but now the big boys in the router business are entering this game and you'll need to keep up. Work on a larger central-site product with more ports. You might even want to consider building a high-speed unit that works at speeds from 192Kbps to T1.

To Cisco: does your entry-level 760-series ISDN router work over leased lines? If not, change it so it does... you'll then become a full competitor to the Yamaha RT100i series.

To (terminal server maker) Livingston: congratulations on finally discovering the Japanese market... various evidence I've seen recently suggests you're getting serious about Japan. However, Ascend has at least a 2-year head start on you with their PRI product, and nobody seems to care how much they've mutated RADIUS with their own proprietary modifications. I don't know if you've got a price advantage for your Portmaster 3 series equipment, but if you don't, you're going to have to find a good set of reasons to convince potential customers why they shouldn't buy Ascend. In addition, one of the best-kept secrets in Japan is that it's possible for ISPs to inexpensively provide 64Kbps and 128Kbps dial-up service to customers using the Portmaster 2 series with an ISDN card. Because it can take so long to get INS1500 into a site in Japan, the PM2 with INS64 lines is an excellent, reasonably-priced alternative for a small ISP trying to provide some 64Kbps and 128Kbps dial-up service. Write up a small Japanese-language white paper on this topic and get it out to your sales force. Include cost comparisons. For a small to medium port count, the PM2 solution will win hands-down over something like the Ascend Max 4000.

To Ascend: What can I say? It's difficult to argue with success. Still, your boxes need at least two things: more CPU and fewer bugs. I have yet to talk to an ISP that uses the full number of PRI ports on an Ascend Max box... apparently there just isn't enough CPU firepower in the box to handle that many calls. Back in the bad old days of 1995, I had first-hand experience with an Ascend Pipeline 400 router that was crashing every 3 to 5 minutes... all because I was foolish enough to think that because the box had 4 ISDN ports, I could actually USE all 4 ISDN ports. Is your tech support department's stock response to "I have this problem, it looks like a bug" still "Upgrade to patch level 5.3b6l2c57, which is today's patch release"? If Ascend hasn't got past the "a new software patch every day!" stage of software quality control, you should get there fast or risk an angry customer base.

To the Japanese government: your gorilla telco, NTT, is trying to become an even larger gorilla. This is getting out of control, assuming that you ever had any control over the situation in the first place. Until you stop accepting fictitious telco breakups, take major steps to deregulate your telecommunications industry, and impose more stringent restrictions on NTT to force down its pricing in its monopoly areas and prevent it from engaging in cross-subsidization, your citizens will continue to pay far more than they have to for telco services and your businesses will not be competitive on an information technology level with businesses in countries with saner telecommunications costs. If you want to know what to do, talk to people in the trenches who have to buy services from NTT, talk to GLOCOM, talk to the consumer, talk to anybody except MPT bureaucrats and NTT managers and board members. You should promote separate facilities-based competition, not the sort of pseudo-competition we have today where everybody else has to purchase rack space in NTT's buildings and buy access to their local loop. You should pass right-of-passage legislation declaring that all conduits, pipes and poles which carry telecommunications lines are a public good which cannot be owned by NTT; no more of this nonsense about TTnet not being able to pull in fiber because NTT owns the poles. Allow voice over IP; if Rimnet can compete with NTT for voice traffic using NTT's own leased lines as the primary conduit of the packets, there's no reason they shouldn't be allowed to do so. Reduce your restrictions on wireless networking to allow license-free wireless equipment, and in particular spread-spectrum equipment, with a range of more than a few hundred meters. 20 kilometers would be nice for starters. And above all else, stop NTT from using its lock on the local loop as a way to prevent competition.

To the Japanese telecommunications-using citizen: are there any public interest groups that lobby on behalf of Joe Citizen regarding Internet and telecommunications issues in Japan? If there aren't, could you make some? Or is this so foreign a concept that it isn't even possible to imagine it? In the U.S. we have the VON coalition promoting voice-over-IP when the telcos tried to squash it. We have the Internet Access Coalition producing studies countering the telco claim that Internet users are clogging up the public phone network. We have a broad coalition of plaintiffs working together in the anti-CDA (communications decency act) court challenge. We have, at least occasionally, the open comments solicitation process from the FCC as a method of the public making sure that the FCC doesn't do something totally stupid. We have, I would like to hope, a process and a culture which help to ensure that the phone companies don't always get what they want... because what your phone company wants isn't always what YOU should want. If citizens don't control this process in Japan, somebody else (the phone company, MPT bureaucrats, politicians, pick your favorite) will.

To the English-speaking Internet consumer in Japan: there are still people out there propogating one of two myths. These are, first, that Niftyserve is the best/only method of obtaining Internet access in Japan, and second, that TWICS is the only method. Whenever you find one of these people, please hit them over the head with blunt objects and tell them to buy either a copy of Internet Magazine (if they read Japanese) or a copy of Computing Japan (if they don't) and to start reading the ads. There are over 1000 Internet providers in Japan today; anybody who still thinks that Niftyserve is a necessity needs an education. And sure, TWICS is still an option, and has been for years, but it isn't the ONLY option.

To NTT: Lay down more fiber. Lay down more copper. Clean up the quality of your copper and your switching equipment in areas where transmission quality is poor. Divert some of your huge piles of money into doing realistic demand projections and investing in switching infrastructure so that you have equipment ready IN ADVANCE of demand. ISP customers should not have to wait 3, 6, or 9 months for your internal procurement wheels to grind to order a new PRI board for your central-site equipment just so the ISP can get another INS1500 line. There shouldn't be an extra 3 month delay in smaller cities to obtain INS64 lines because your copper isn't clean.

** No fate but what we make **

In June 1994, in an essay which appeared in Computer-Mediated Communication magazine, I wrote these words after attending the 1994 Tokyo Business show:

"I was keeping my eyes out for the magic word 'Internet' in katakana, but I saw it almost not at all."

Today, nearly 3 years after I wrote those words, it's impossible to NOT see the word "Internet" at any technology show in Japan. We've gone from 2 providers of dial-up service in the entire country, one selling only UUCP at Y30/minute (IIJ) and one only providing VMS shell accounts (TWICS, which started offering PPP in 1995), to over 1000 ISPs in hundreds of locations. We now have nationwide call-routing services available from the various NTT competitors which allow end users anywhere in the country to connect to a variety of Internet providers for rates as low as Y9/minute... still a heavy surcharge, but better than the bad old days when some users were going so far as to use international callback systems to reach their Tokyo-based Internet providers, since the $1.00/minute or so that they paid to route the call to the U.S. and back was still lower than NTT's domestic long-distance charges to call Tokyo. We've seen the total Internet bandwidth into Japan increase from about 1 Mbps total in late 1994 to multiple T3 (45 mbps) circuits today. We've seen leased line and dial-up prices slashed and burned in price wars which have lowered the price (and, frankly, almost certainly lowered the average quality) of Internet service to the end user. And for the most part, with one or two notable exceptions, the government has steered clear of the censorhappy attempts to control content which the governments of the U.S., Singapore, Germany, and other nations have disgracefully indulged themselves in. Internet life in Japan is certainly better than it was in late 1994, and I'm happy to have played my part in helping it to grow.

Yet despite the improvements, I'm haunted by feelings that life could be better, and I find myself always asking "what if..." questions. What if leased line circuits in Japan had a price per unit bandwidth per unit distance comparable to those in the U.S? What if opening an Internet exchange were as simple as providing the housing, the electrical power, and an air conditioner? Would we see a new collection of IXes pop up in unexpected locations... Sapporo IX, Kyuushu IX? What if unlicensed spread-spectrum wireless devices were allowed to transmit at 1 watt of power instead of the 10 mw I keep seeing in the spec sheets? What if there were consumer-friendly and competition-friendly interconnection regulations, rigorously enforced? What if the government were more interested in promoting competition and new methods of communication than in protecting NTT's interests? What if anybody could order an international line, or set up a satellite transmission system, without needing at least a few million dollars in working capital and various permissions from the government? What if NTT were REALLY broken up... into 8 or more separate local loop companies ("baby NTT's"?) and several long-distance-only companies, with a consumer-friendly watchdog organization appointed to prevent collusive practices between the new corporations? What if it didn't cost US$700 to install a phone line? What if voice-over-IP were legal? What if I had legal recourse to force faster action when NTT tells me it's going to take 9 months to put in an ISDN line? What if telecommunications regulations were made on a regional or city-by-city basis instead of by one Tokyo-cental organization? Might we see some cities gaining reputations as "telco havens" for business due to their efforts to promote competition and push down costs?

There are some things, it's said, that man was never meant to know. And, wonder as I might, I don't see much chance of any of my questions being answered any time this century. I hope, for once, that I'm dead wrong.


Author's bio:

After earning his M.S.E.E. from Cornell University (Ithaca, New York), Bruce Hahne came to Japan in the fall of 1993. After one year of grinding out C++ code for Mitsubishi Electric and continued frustration with the poor quality and low availability of Internet service in Japan, he escaped the clutches of corporate Japan in 1994 to help create the Internet service provider Global OnLine Japan K.K., where he served as V.P. Technology until 1996. He is presently finishing a brief stint as head engineer for Business Network Telecom K.K. Where he goes after that is anyone's guess, though hopes are high for maintaining some involvement in Asia-Pacific Internet and communications development. He has recently created a new company, ISP Solutions Inc., whose goal is to provide cutting-edge hardware and software technologies to ISPs worldwide.

---------------------------------------------------------------------------

References and resources:

Publications:

- Computing Japan's web site is at www.cjmag.co.jp - Internet Magazine's publisher, Impress, is at www.impress.co.jp - Boardwatch magazine is at www.boardwatch.com - The Computer-Mediated Communication archives are at www.december.com/cmc/mag/archive/index.html

Providers and telcos:

- Tokyo Internet is at www.tokyonet.ad.jp - IIJ is at www.iij.ad.jp

- TWICS is at www.twics.com

- Rimnet is at www.rim.or.jp

- NTT is at www.ntt.co.jp

- News items about Direct Internet are at:

www.pcronline.com/pcnews/1127/010.html

www.dbsdish.com/news3/news1773.html

www.tele-satellit.com/listserver/tags1/sat-nd/msg00205.html

- OZemail, provider of voice-over-IP services in Australia, is at www.ozemail.com.au

- KDD is at www.kdd.co.jp

- IDC is at www.idc.co.jp

- ITJ is at www.itj.co.jp

- DDI is at www.ddi.co.jp

- TWJ is at www.telewaynet.ad.jp

- JCC is at www.jcc.co.jp

Equipment and product suppliers and manufacturers:

- Livingston Enterprises is at www.livingston.com

- Ascend Communications is at www.ascend.com

- Cisco Systems is at www.cisco.com

- US Robotics is at www.usr.com and www.usr.co.jp

- Rockwell Semiconductor is at www.rockwell.com

- Yamaha is at www.yamaha.co.jp

Organizations:

- WIDE, Japan's academic Internet, is at www.wide.ad.jp

- GLOCOM's home page is at www.glocom.ac.jp

- Japan's MPT is at www.mpt.go.jp

Other:

- The APRICOT conference home page is at www.apricot.net

- The voice-over-net coalition is at www.von.com

- The Internet Access Coalition is at www.internetaccess.org

- The home site for the anti-CDA court challenge in the U.S. is at www.aclu.org/issues/cyber/hmcl.html

- Various fact-checking and background research for this essay was done using the Altavista search engine at www.altavista.com.

- The Newsbytes Pacifica archives at www.nb-pacifica.com/headlines/archives.html

and the GLOCOM Netizen mailing list archives at www1.glocom.ac.jp/Netizen/archive-e/ provided reference material on the NTT "breakup" announcements of late 1996.

- Thanks to Stephen Anderson of GLOCOM for comments on a preliminary draft of this essay.

--------------------------------------------------------------------------

Disclaimers: With the exception of Business Network Telecom, I don't work for any of the companies listed above. All opinions are mine; any resemblance to the opinions of others, or to anything vaguely resembling sanity, is completely coincidental. These opinions are free, so you get what you pay for, and anybody running out and investing in somebody's stock just because of something I wrote here is not allowed to come yelling to me when a competitor comes out with a vastly superior product next month. Send factual corrections to my email box, flames to /dev/null.

Legal matters: This essay is copyright (c) 1997 by Bruce M. Hahne. Republication or redistribution of this essay for non-commercial purposes, including noncommercial archiving on web, ftp, and gopher sites, is permitted and encouraged so long as this copyright notice and the contact information below is maintained. Although it's not required, it would be nice if you'd drop me email to let me know where you republished it. Republication or redistribution in any form for commercial purposes is not permitted without my permission; please contact me first and we can discuss your plans.

Contact information:

Author's present address: hahne@giganet.net

Author's permanent address: hahne@acm.org

Affiliations: EFF (Electronic Frontier Foundation), CPSR (Computer Professionals for Social Responsibility), ACM (Association for Computing Machinery), ACLU (American Civil Liberties Union).

--------------------------------------------------------------------------

GLOCOM homepage [J] [E]