[25] in Commercialization & Privatization of the Internet

home help back first fref pref prev next nref lref last post

Was a "big Internet" needed to make TCP/IP useful?

daemon@ATHENA.MIT.EDU (John Gilmore)
Tue Oct 23 06:25:59 1990

Date: Tue, 23 Oct 90 02:45:00 PDT
From: gnu@toad.com (John Gilmore)
To: com-priv@psi.com

Craig Partridge wrote:
> Beating them over the head with a problem they're trying to fix seems
> (a) ungrateful,

I'm pleased that when the War Dept. hired people to build them networking
technology, the hired people made it more generic than just for war.
Thanks, everyone!

>                 and (b) likely to make transition planning harder because
> by forcing the Feds to make explicit policy, you restrict their flexibility.

I see no transition plan.  Everyone complains about the weather, but
nobody does anything about it.  The plans I hear are all to spend more
money on more bandwidth for more people to slime in violation of the
rules, which get looser and looser.  We seem to be in transition, but
somebody left the gearshift in forward, not reverse!

>     Further, the comment about the Feds prefer to taking it from us and
> then give it away is inconsistent with past NSF history on networking.

I was speaking of a generic tendency in government; NSF may have it to 
a lesser degree, but note that the mere existence of tax-funded "national
science" research demonstrates the point.  After NSF made CSnet go
self-sufficient, did it get back its investment, like any commercial
investor would've done?  No, it went back to the taxpayers' pockets to
fund its next projects!

>                                                    problem was, no
> one else existed in the early 1980s willing to fork over the $$ necessary
> to make a big Internet happen.

It is unclear to me whether the "big Internet" was necessary to make a
large commercial network viable.  Basically DARPA-funded BSD Unix came
out in the early '80s with TCP/IP "for free", the most successful
vendors shipped it with their computers.  While I'd credit their
interaction with the usenet/internet news/mail system with ~5% of Sun's
success, I don't think a big free net of high speed leased lines caused
any more workstations to be sold.  (Sun itself wasn't even on the
Internet til '85).  Given a large installed base of TCP/IP machines,
the incentives to interconnect them were there already, and indeed
Telebit did (and does) a brisk business in 19.2kb uucp
interconnection.  But high prices of dedicated routers, and a year or
two's stall on SLIP versus PPP, pushed anyone who wanted TCP onto the
Internet, and instead of pushing inappropriate traffic off, NSF
expanded the bandwidth.  And today they're doing the same thing again,
running up the deficit for T3 bandwidth, even though now there are
useful pay-your-own-way alternatives.

BTW, we are seeing the same 1-2yr stall on ISDN, another "SLIP-like"
technology which would make "roll your own Internets" almost as easy as
uucp.  Every SPARCstation-1, 1+, SLC, and IPC has an ISDN chip in it
that could hook it up cheaply to wide area 128kb/sec networking.  But
one particular person in Sun Networking Software blocked the writing of
SLIP or PPP drivers for it, preferring to wait for a more hardware
intensive solution involving DMA'd X.25 over ISDN.  Since no software
supported ISDN, the hardware people removed the capability.  It's in
the chip, but the passive components required to bring it to a
connector are gone now.  There would've been 50,000 ISDN end user
machines capable of plug-and-play TCP/IP over ISDN by now, many at
under $5000, had this not been botched.  And Pac Bell would've found a
way to fund nonlocal ISDN if they had a thousand California customers
demanding it for their Suns, too.  (It's currently held up behind
Caller-ID, which the PUC has put on hold.  So even where you can get
ISDN it only works in your central office, not to any other CO's, and
there is no schedule for fixing this.)  I can't blame this one on NSF
:-) but I wonder if we'd really need a massively expensive gigabit
network if we had tens of thousands of 64 or 128kbit links that
actually went to somewhere that somebody needed communication.  I
somehow doubt that Cygnus Support will be on the giganet, though we'll
pay for it.

	John Gilmore

home help back first fref pref prev next nref lref last post