Microcom

» see microcom.com as it lived, on the wayback machine

~ the eulogy ~

Microcom was founded in 1980 by James M. Dow, who came out of Data General, and it sold modems out of Norwood, Massachusetts for seventeen years, going public seven years in. It was never the biggest name in the category. Hayes owned the command set and USRobotics owned the enthusiast market. Microcom sat behind both of them for its entire life.

What it had instead was a protocol. In the mid-eighties every modem vendor was trying to differentiate against the flood of cheap Hayes 1200 clones, and the obvious route was faster modulation, which required an expensive microcontroller and a lot of engineering. Microcom went sideways. It wrote a family of packet protocols called MNP that a modem could run on the cheap microcontroller it already had, and that turned an error-prone phone line into a clean one.

The hardware followed the protocol out. The AX/1200 and AX/2400 carried MNP 4 error correction on top of otherwise ordinary V.22bis modulation, so two Microcom modems talking to each other produced a connection with no corrupted characters in it. The AX/1200c and AX/2400c added MNP 5, where the modem compressed the data before sending it and the effective throughput went above the line rate. The AX/9624c pushed to 9600 bits per second using a variation on V.29 and Rockwell's fast training, at a lower price than the proprietary 9600 boxes everyone else was shipping.

Later the line broadened into price segments with names that told you where each one belonged. DeskPorte sat on a desk and the TravelCard units went into a laptop's PC Card slot. Alongside the hardware there was a software business, including the Carbon Copy remote control package, and a rack product called HDMS that early internet providers bought for their first banks of dial-in ports.

~ the dossier ~

~ what mnp actually did ~

The levels were graded by how hard they were to implement rather than by how well they worked. MNP 1 was deliberately trivial, simple enough to drop into existing modems with no hardware change, while MNP 4 gave much better throughput at the cost of memory, which was the scarce resource in a modem where forty bytes of buffer was a normal allocation. The one that mattered commercially was MNP 5. Compression inside the modem meant a 2400 bit per second line could move text faster than 2400 bits per second, and that was a number you could print on a box.

Then Microcom gave the protocols away. MNP 1 through 5 went to ISO for standardization and became available to everybody, and within a few years most modems on the market could negotiate MNP whoever had built them. The company kept developing further levels, including MNP 10 and MNP 10EC for cellular links, built jointly with Rockwell. This was not a mistake. Microcom made money licensing to competitors and it made its own modems interoperable with theirs, which is the only condition under which a small vendor survives against Hayes. What it could not do was make MNP a reason to buy a Microcom modem specifically, once every modem had it.

~ the init string ~

I got online in 1998 on a 14.4k modem and worked up to 56k over the next few years. Every one of those boxes had a line on the packaging listing the error correction and compression it supported. V.42bis was on it. So was MNP 5. Those four letters were on the spec sheet of everything I bought and everything I looked at and did not buy, and it never occurred to me that they abbreviated the name of a company, let alone a company in Massachusetts that had been sold the year before I got my first modem.

By then MNP had mostly lost to V.42 and V.42bis, which were the ITU versions of the same ideas and performed better. It survived as the fallback, the thing your modem reached for when the far end could not do V.42, which is why it stayed printed on the box long after it stopped being the feature anybody chose on. That is a strange kind of afterlife for a protocol, and it lasted longer than the company did.

~ what compaq wanted ~

Compaq announced the deal on 10 April 1997, a cash tender at $16.25 a share for roughly $280 million, on a merger agreement dated the day before. The offer expired at midnight on 13 May and the acquisition completed on 14 May 1997. The stated logic was to pair Microcom's modem and remote access work with Compaq's Windows NT server business and sell connectivity end to end.

It lasted about twenty months. By January 1999 Compaq was scaling its networking business back, and the modem line went with the retreat. Microcom's name disappeared from the market it had spent seventeen years in, while the protocol it had written stayed in the firmware of every modem still shipping.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »