Re: Overseas connections a waste of time???

[turnbull@example.com (Stephen J. Turnbull)]

>>>>> "Dennis" == Dennis McMurchy <denismcm@example.com> writes:

    Dennis> On Wed, 9 Oct 1996, Stephen J. Turnbull wrote:
    >> As far as I can figure out, the entire SINET community,

    Dennis>   My (of course hopelessly out-of-date) "Directory of
    Dennis> Electronic Mail Addressing & Networks" has an entry for
    Dennis> SINET = Slovenia IP network.  Great though my faith in the
    Dennis> printed word is, I somehow suspect this is not the SINET
    Dennis> you're referring to.  JUNET, of course, I've heard of, but
    Dennis> what is SINET?

"Scientific Information NETwork."  JUNET you've heard of, but it's too
late.  It's dead, just like the ARPAnet.  Of course, it lives on, just
as the ARPAnet does (if you talk with your nameserver much, you'll
still see IN.ARPA.123.123.123.123 addresses...).

    >> This wasn't so much due to primitive infrastructure as to
    >> punitive pricing policies by the providers, encouraged by your
    >> friend and mine, Yuseisho, which wanted to prove that Internet
    >> growth in Japan could be held below economic growth.  (That's a
    >> very bitter, bitter joke---I hope.)
 
    Dennis>   It's far from being a joke, I'm afraid.  The Yuseisho is
    Dennis> a powerful component of the bureaucratic dictatorship that
    Dennis> effectively controls this country.

I'm aware of the importance of Yuseisho.  I specifically had in mind a
prediction Yuseisho made a few years back that by 2000 there would be
room for only two or three ISPs of NTT's size as of 1990.  The "bitter
joke" was that Yuseisho would intentionally keep the Japanese Internet
so small that that prediction would come true.  This I don't believe,
but...

    Dennis> The Yuseisho and NTT together have done so much to retard
    Dennis> the healthy development of the Internet in Japan that it

... different viewpoints exist about what's "healthy," so ...

    Dennis> is almost hard to believe that this was just gross
    Dennis> bureaucratic bungling (as it most probably was) and not
    Dennis> the result of a deliberate policy.

... it *is* the result of a deliberate policy.  Japanese bureaucrats
(which term includes almost all senior managers of listed companies,
as far as I'm concerned) believe in *controlled* growth of *orderly*
markets.  "Controlled" in Japan means that every detail is watched by
some bureaucrat or manager, not that policies are in place so that the
mountains move for Mohammed and the molehills take care of themselves.
Thus the preference for a small number of big companies, and measures
taken to control what growth occurs.

    Dennis>   Apparently NTT (having recently become aware of the
    Dennis> existence of computer networks) is now set to become the
    [snip]

I don't think so.  NTT America has been a pretty big player on the net
for several years.  Their response to the need for Internet-based info
in the Kobe quake crisis was quite impressive, much more so that the
whole crowd of Tokyo-based providers, not to mention the !@#$%
government, which wouldn't even answer phone calls from the Japanese
consulate in Seattle.  What they recently became aware of is that
Japan needs access at the private individual and small business level
to be internationally competitive.

    [snip]
    Dennis> leaving the Yuseisho-NTT with an effective monopoly of the

NTT *will* be broken up in the near future.  What this implies for the 
ISP market is not clear; the so-called "new common carriers" will be
in a position to compete to some extent.  How great an extent depends
on whether MITI and the MOF can put the fear of God and/or Mickey
Kantor into the heart of Yuseisho.

    Dennis> ISP market.  This leaves the whole show in the hands of
    Dennis> the retarded, nasty and self-serving bureaucrats, to
    Dennis> manipulate and control largely as they see fit.

They're not "retarded," they're the best and brightest.  Seriously.
"Nasty?"  Most of the (policy level) Japanese bureaucrats I've talked
to have been quite charming.  "Self-serving?"  Of course.  Who isn't?
But they also believe in what they're doing, I think, in many cases.
"As they see fit," unfortunately, is right on the nose.  There is no
market-based or election-based check on their power.

    Dennis> To my mind this bodes ill for the future of the
    Dennis> Internet in Japan.

Agreed.  Whatever it is that the bureaucrats decide is good for Japan, 
we can be pretty sure they won't be "ten years early."

>>>>> "Craig" == C Oda <craig@example.com> writes:

    Craig> The connection speeds inside of Japan are much higher than
    Craig> the connection speeds to places outside of Japan for must
    Craig> ISPs in Japan.  The instrastructure among Japanese
    Craig> providers has improved recently.

    Craig> I am connected to SprintLink, but my speed to someplace
    Craig> like Tokyo Internet is quite fast:

PING www.tokyonet.ad.jp: 64 byte packets
64 bytes from 202.239.61.53: icmp_seq=0. time=52. ms

PING www30.netscape.com: 64 byte packets
64 bytes from 207.12.240.31: icmp_seq=0. time=222. ms

PING gol.com: 64 byte packets
64 bytes from 202.243.48.4: icmp_seq=2. time=437. ms

    Craig> Traceroutes often identify bottlenecks to a Japanese site
    Craig> at places not related to NSPIXP or WIDE.

    Craig> For some reason, traffic from Japanese sites to gol is
    Craig> being routed through California in the U.S.  Route 5 is the
    Craig> Sprint switching point in stockton California.  The return
    Craig> times jumpt from about 50 to 220ms when the signal leaves
    Craig> Tokyo.

Well, GOL's internal network seems stuffed half the time.  Note how
the response time jumps between WIDE and GOL (nodes 9 and 10,
respectively).

traceroute to gol1.gol.com (202.243.48.4), 30 hops max, 40 byte packets
 1  130.158.99.254 (130.158.99.254)  2 ms  1 ms  1 ms
 2  RD03.cc.tsukuba.ac.jp (130.158.1.82)  4 ms  3 ms  3 ms
 3  fuyou-fddi.ric-tsukuba.ad.jp (192.70.180.254)  4 ms  3 ms  3 ms
 4  new-tsukuba.bb.sinet.ad.jp (150.99.3.1)  7 ms  10 ms  7 ms
 5  new-nacsis.bb.sinet.ad.jp (150.99.105.1)  22 ms  8 ms  19 ms
 6  nishi-chiba.sinet.ad.jp (150.99.90.1)  8 ms  8 ms  8 ms
 7  otsuka.sinet.ad.jp (150.100.126.1)  15 ms  46 ms  11 ms
 8  otsuka-gate.sinet.ad.jp (150.100.1.4)  13 ms  13 ms  13 ms
 9  cisco12.tokyo.wide.ad.jp (133.4.3.4)  18 ms  26 ms  15 ms
10  gol.nspixp.wide.ad.jp (202.249.3.42)  138 ms  145 ms  136 ms
11  202.243.49.5 (202.243.49.5)  290 ms  459 ms  302 ms
12  gol1.gol.com (202.243.48.4)  228 ms  133 ms *

Same thing happens with TWICS, though:

traceroute to pollux.twics.com (192.135.222.11), 30 hops max, 40 byte packets
[snipped, essential same as above]
 9  cisco12.tokyo.wide.ad.jp (133.4.3.4)  17 ms  16 ms  15 ms
10  gsl.nspixp.wide.ad.jp (202.249.3.48)  192 ms *  349 ms
11  * * *
12  gsl-Twics.gsl.net (204.59.68.42)  233 ms  205 ms *
13  pollux.twics.com (192.135.222.11)  204 ms  218 ms  308 ms

Sheezus, lookit them dropped packets!  This stupidity does *not* happen
with *any* of the US sites I've tried:

traceroute to ecolan.sbs.ohio-state.edu (128.146.137.2), 30 hops max, 40 byte packets
 1  130.158.99.254 (130.158.99.254)  2 ms  1 ms  1 ms
 2  RD03.cc.tsukuba.ac.jp (130.158.1.82)  5 ms  3 ms  3 ms
 3  fuyou-fddi.ric-tsukuba.ad.jp (192.70.180.254)  4 ms  8 ms  3 ms
 4  new-tsukuba.bb.sinet.ad.jp (150.99.3.1)  5 ms  7 ms  7 ms
 5  new-nacsis.bb.sinet.ad.jp (150.99.105.1)  12 ms  9 ms  9 ms
 6  nacsis-gate3-F2/0.sinet.ad.jp (150.99.90.3)  8 ms  8 ms  9 ms
 7  sl-stk-7-S4/0-4xT1.sprintlink.net (144.228.47.113)  129 ms  123 ms  130 ms
 8  sl-stk-1-F0/0.sprintlink.net (144.228.40.1)  125 ms  124 ms  127 ms
 9  * core4-hssi-1.SanFrancisco.mci.net (206.157.77.65)  129 ms  126 ms
10  borderx1-fddi-0.WillowSprings.mci.net (204.70.104.20)  170 ms  380 ms  211 ms
11  borderx1-fddi-0.WillowSprings.mci.net (204.70.104.20)  221 ms *  170 ms
12  * cicnet.WillowSprings.mci.net (204.70.104.102)  170 ms *
13  dgf-fddi8-0.chicago.cic.net (192.217.0.4)  170 ms  171 ms  170 ms
14  dgx-fddi5-0.chicago.cic.net (131.103.1.34)  170 ms  170 ms  168 ms
15  osu-hssi1-0.columbus.cic.net (131.103.22.130)  182 ms  180 ms  185 ms
16  kc2-fddi3-0.net.ohio-state.edu (192.148.245.5)  178 ms  180 ms  179 ms
17  se2-atm4-0s10.net.ohio-state.edu (164.107.1.193)  183 ms  178 ms *
18  * ecolan.sbs.ohio-state.edu (128.146.137.2)  184 ms  183 ms

Traceroutes to wuarchive.wustl.edu (Washington U, St Louis),
tsx-11.mit.edu (Boston), and ftp.cc.gatech.edu (Atlanta) give similar
results; all under 200ms at the moment.  With the air RT to Boston
clocking in at just about 20,000 km, that's about 1/3 speed of light.
Not bad at all :-)

GOL to the US is quite fast, sufficiently so that although my telnet
connections to GOL are unusably slow (I've completely given up Usenet
except for the comp.* groups that my Institute's server feeds; last
time I checked Tsukuba-dai's server did *not* correctly implement
NNTP!!), I can often get triple the throughput (plus a 50% reduction
in dropped connections) by FTPing US --> GOL and GOL --> Tsukuba-dai.

    Craig> How does all this relate to Linux in Japan?  Well, I'm very
    Craig> satisfied with the connection of my Linux box to the
    Craig> Internet.  I home box has 128Kbps ISDN with STAC
    Craig> compression.  However, I almost always only use one B
    Craig> channel at 64Kbps.  For me, it is fast enough.

This is very nice, but my office box *does* get effective rates of
1Mbps (for FTP transfers) to sites on the SINET backbone (say
ftp.tohoku.ac.jp or ftp.lab.kdd.co.jp), but often averages 14.4 kbps
or so to gol1.gol.com, and it *doesn't* go via Stockton, thank-you-
very-much, it goes via a ring of Ciscos in Tokyo.

My guess is that in Japan for the next couple of (maybe many) years,
it's going to matter very much where on the net your box lives.  And
for many folks, getting that ISDN connection is not going to make
sense because they don't have a primary connection point (ie, work) as
Craig does (and I do, for that matter).

By the way, the !@#$% gateway from Tsukuba-dai to the rest of the
world was down when I started this message (I got the traceroutes
later).  If Dennis is right, and NTT is going to be able to maintain
an effective monopoly over Internet services, we're going to continue
to see corporate and academic consortia rolling their own networks
over leased lines, and that kind of crap will continue for the
indefinite future.  Some people will get excellent 24-hours-a-day
service, and others will get stuffed.

:-(

Steve

-- 
                           Stephen John Turnbull
University of Tsukuba                                        Yaseppochi-Gumi
Institute of Policy and Planning Sciences  http://turnbull.sk.tsukuba.ac.jp/
Tennodai 1-1-1, Tsukuba, 305 JAPAN                 turnbull@example.com
-----------------------------------------------------------------
a word from the sponsor will appear below
-----------------------------------------------------------------
The TLUG mailing list is proudly sponsored by TWICS - Japan's First
Public-Access Internet System.  Now offering 20,000 yen/year flat
rate Internet access with no time charges.  Full line of corporate
Internet and intranet products are available.   info@example.com
Tel: 03-3351-5977   Fax: 03-3353-6096