The Little Garden

Source

Below the fold is the story of how I got a full-time Internet connection at my apartment 32 years ago next month, and the incredible success of my first ISP.

The reason I'm now able to tell this story is that Tom Jennings, the moving spirit behind the ISP has two posts describing the history of The Little Garden, which was the name the ISP had adopted (from a Chinese restaurant in Palo Alto) when I joined it in May 1993. Tom's perspective from the ISP's point of view contrasts with my perspective — that of a fairly early customer enhanced by information via e-mail from John Gilmore and Tim Pozar, who were both involved far earlier than I.

Jennings starts his story:

Once upon a time, three little businesses wanted a connection to the ARPAnet/internet. The year was 1990 or 1991. John Gilmore, John Romke[y], and Trusted Information Systems (TIS) split the $15K or so it took to get a leased-line and 3COM Brouters to Alternet, with what today you'd call fractional T1. An additional 56K leased line and Brouter brought the 'net up to Gilmore's house, Toad Hall, in San Francisco.

The three little businesses were Cygnus Support (John Gilmore), Epilogue Technology (John Romkey) and Trusted Information Systems (Steve Crocker). AlterNet was run by Rick Adams, whom Wikipedia justly describes as an "Internet pioneer". He founded UUNET Technologies:

In the mid-1990s, UUNET was the fastest-growing ISP, outpacing MCI and Sprint. At its peak, Internet traffic was briefly doubling every few months, which translates to 10x growth each year.

John Gilmore, a truly wonderful person, had many friends. So what happened was:

As time went on, friends of theirs wanted in on this rare and exciting 'net connection, resulting in Tim Pozar putting an old PC running Phil Karn's KA9Q/NOS program, an amateur radio router capable of TCP/IP, onto Toad Hall's ethernet. Tim installed a pair of modems, then dialed in once and stayed connected 24 hrs/day (Pacific Bell never said you couldn't do that...)

Once Tim showed that it was possible, this idea took off:

Eventually the NOS box was full, and more friends wanted in, but everyone was too busy to deal with the hassle.

Somehow, in September 1992, Pozar and Gilmore and I worked out a deal where, I would maintain the thing, collect money to build more NOS boxes and contribute to the monthly Alternet bill, install more people, and get (1) a free connection to the internet and (2) a slice off the top after it exceeded N connections.

By that December, there were enough connections in place that I was pocketing $420/month. By March 1993 there were 11 modem-connected members (as we fancied ourselves).

In 1989 Gilmore had co-founded Cygnus Support, whose tagline was "Making free software affordable". TLG got started in August 1990 with the three businesses' nodes on a 56K leased line. One was at Cygnus first office in an apartment complex on University Avenue in Palo Alto. Gilmore and other Cygnus employees had apartments there, so they used 10BASE2 coaxial cable Ethernet to distribute the Internet around the complex. Gilmore notes that they used "nonstandard thin 50-ohm coax in the expansion joints across the driveways when needed". Pozar notes that they paved over the coax!

Gilmore was paying more than $300/mo for modem phone lines supporting the Alt Usenet groups, and realized that for less than that he could have a 56K line from Cygnus to his basement in SF. That led to Pozar and Rich Morin's Canta Forda Computer installing the old PC and becoming the first to use the permanent local call idea.

I knew Gilmore from the early days of Sun Microsystems (he was employee #5), so I first found out about the Point of Presence (PoP) in his basement in late 1992 and really wanted to join in. Alas, there was a snag — the reason the idea worked was that local phone calls were free. From my home in Palo Alto to Toad Hall was a toll call, making it impossibly expensive. But in May 1993 I found out about the PoP on University, 8 blocks from my apartment.

I purchased:

• A phone line from Pacific Bell at the PoP.
• A second phone line at my apartment.
• A pair of Zyxel 9.6Kb modems, one for the PoP and one for my apartment.

SparcStation SLC

If memory serves, it cost $250 installation fee and $70/month, and Tom Jennings helped me plug in one of the modems at the University PoP. I already had two SparcStations, a SparcStation SLC with an external SCSI hard disk I bought on Sun's employee purchase program, and a SparcStation 1+, the prizes Steve Kleiman and I won in an internal "Vision Quest" at Sun. My apartment was open-plan and the 1+'s fans were too noisy to let me sleep, but the SLC was fanless and could be on-line continuously. The SLC, the hard disk and the modem sat on a conveniently large window ledge. There was a wired Ethernet connection from the window ledge to the desk. When I say "wired" I mean that it ran on the apartments internal phone wires, but the distance was short enough that it worked.

SparcStation 1+

This setup was remarkably reliable. If the call dropped, the SunOS SLIP software automatically re-dialled it. I have no memory of problems with it; I think the only times it was down were when I upgraded the modems as faster ones became available, or when I put the whole system on an Uninterruptible Power Supply. It may have been then that I noticed it had been up over 500 days. I didn't really need the UPS, Palo Alto's municipal utilities are also very reliable.

As I recall it ran happily until I passed the apartment on to my step-daughter's family in summer 2000. Seven years of impeccable service. By that time I was working on the LOCKSS program at Stanford, and we had DSL service from Stanford IT. So I went from an ISP with great tech support to an ISP with great support. Then as I relate in ISP Monopolies in September 2001 Palo Alto's Fiber-to-the-Home trial went live and I had 10Mbit bi-directional fiber with great support from Palo Alto Utilities. Since the trial ended our ISP has been Sonic, first over 3/1Mbit DSL and now over gigabit fiber. So we are really used to having great support from our ISP.

TLG was an astonishing success. From something like $2000/month in December 1992 it grew "an average of 12% per month from Jan94 through July96" when it had "a monthly gross of about $125,000.00 until:

Luckily we were bought by Best Internet Communications, Mountain View; they had money, marketing, and a non-burned-out management; we had a solid locked-in customer base and positive cash flow.

Best turned out to be a pretty good ISP too.

Jennings' explanations for TLG's success are interesting. First, technical competence:

Edgar Nielsen almost single-handedly built the technical infrastructure that TLGnet ran on. He designed much of the network and routing structure, all of the security (with some help from Stu Grossman), wrote a complete, queryable, shared and remotely-accessible database (included every single modem, router, wire, cable, customer, IP (domain names and IP address allocations), and logical link) in standard and portable tools, installed equipment, built and maintained our unix boxes, put SNMP on every single node (hundreds) and automated the entire ISP technical infrastructure from one end to the other. I doubt many small to mid-size ISPs today have the things Edgar wrote by 1995.

Second, good HR:

Another thing of crucial importance to me, and to Deke, Edgar and a lesser extent Gilmore, was hiring from our local communities; we hired principled people, punk and queer writers and organizers, and trained and paid them -- pay in scale with effort. Total staff turn-over in three years was probably 20; peak staff was 12. Some 10 of them started out at $8.00/hr, unskilled, ended up with $30,000 salary a year later [1994-1996], and stayed in the industry (at prevailing pay). (And we provided health insurance too. Deke being damned Wobbly may have had some small effect.)
...
we treated our staff well, gave them credit for work done, paid them actual money, gave raises and bonuses (upon sale of the business, even some fired employees got small bonus checks). TLGnet wouldn't have existed without its talented staff!

Third, an innovative business model starting with their terms and conditions:

TLGnet exercises no control whatsoever over the content of the information passing through TLGnet. You are free to communicate commercial, noncommercial, personal, questionable, obnoxious, annoying, or any other kind of information, misinformation, or disinformation through our service. You are fully responsible for the privacy of, content of, and liability for your own communications.

Jennings explains the business model:

• Concentrate on bulk, fulltime internet access (leased-line and Frame Relay)
• Keep prices low by providing connectivity only
• Unrestricted use of TLG connectivity
• Encourage resale and vertical-market services
• Full, up-front disclosure of all pricing
• No lock-in contracts
• Unbundle installation costs and eliminate padding
• Full technical disclosure of technical information

Essentially, other ISPs restricted use and resale of their connections, in a sort of zero-sum approach. By concentrating on bulk connectivity we at once created a market for our customers to provide the vertical services we didn't want or couldn't afford to provide, and built a hard-to-beat solid rep that for a long while locked out direct competitors to our core business; having our prices online and breaking down the leased-line costs and equipment gave us a major one-up economically, technically, and in credible reputation over nearly all other ISPs, big or small.

The result was:

Some thought us insane; but in fact our customers didn't "compete" with us, they provided vertical services we couldn't or wouldn't (I guess we did have a business plan). And in fact we set further standards of behavior and policies that other ISPs, including MCI and SprintLink, were obliged to match. Though some, like Alternet and PSI, never did; they skimmed the high-end deep-pockets customers, and we got all the new growth.

Gilmore writes:

I would add to the "Busines Model" discussion, that communication costs per-bit dropped dramatically with volume. When you upgraded from 56k bit/sec leased lines to T1 (1,500k bit/sec), you got 24x the bandwidth but it only cost about 4x as much. An upgrade to T3 (45 megabit) provided 30x the bandwidth of a T1, and didn't cost anything near to 30x as much. So, as your traffic volume grew because you were adding more and more customers, the cost of your basic connection to the rest of the Internet got significantly cheaper (per bit). That economy of scale meant that ISPs who grew could keep affording to upgrade their backbones to handle the traffic growth. Every ISP knew, or figured out, this economics, and they all depended on it. Remember, this was back when there were 2000 ISP's in the US, mostly local ones. (About 30 of them were getting their Internet service from TLG when we sold it to Best.)

There is a fascinating October 29 1996 interview entitled Tim Pozar and Brewster Kahle CHM Interview by Marc Weber. The first part of the interview is all about TLG. In it Brewster Kahle sums up the story (I cleaned up his stream of conciousness a bit):

it took six months of a full-time person to get us on the DARPA net in 1985 ... but The Little Garden basically made it so that any old person [could connect] and more than that not just themselves but ... enabling other people to create their own ISPs and I don't know there are 400 ISPs now in the Bay Area in large part because of The Little Garden.