Letters to the editor: “Ethernet’s secret source”

Also: the digital lifestyle, Explorer bugs

Not a flop

I disagree with Kevin Tolly including Arcnet in his list of flops. Arcnet was token passing over star-connected coax and worked quite well if you kept the number of stations on a segment to a few dozen. In a 150-station law firm, I put four Thomas-Conrad Arcnet cards in each of two primary NetWare servers, with a 10BaseT backbone between the servers. It worked very well.

Thomas-Conrad Arcnet was installed before I joined the firm, but the manageable hubs were stacked on top of one another in unventilated closets and were running so hot that the rubber feet had melted. Yet even then, it was reliable. The only glitch was that nobody had bothered to track the node numbers — Arcnet had 8-bit addressing, set with a DIP switch on the card. Users would get knocked off the network when someone else with the same node number booted up on the same network segment. I solved that problem by using a NetWare command to show login name, network branch number and node number, piping the results to a Sort filter, so I got lists of node numbers for the whole network, by branch. That enabled me to set all node numbers to unique values and ended that problem. Ethernet with unique MAC addresses is a giant step forward in that aspect.

Later I used the Arcnet cabling with recycled TCNS cards and a hub to demonstrate 100mb performance to management, convincing them to invest in Cat 5 wiring, and we rolled the whole firm into switched 100BaseT. Arcnet was excellent in its time. TCNS was just too expensive for most IT departments.

Fred Wagner

Document systems specialist

City of Long Beach, Calif.

Kevin Tolly’s column brought back fond memories of some of the columns he wrote (and I responded to) back in the mid-to-late 1990s as he tried desperately to keep token ring going in the face of Ethernet inevitability. All the buzzwords were there, including the ever-popular “non-deterministic performance” and “rapid performance degradation with traffic increases.”

At that time I ran a large token-ring shop after running a large Ethernet shop, and I’m afraid I have to concur with the Boggs paper that “Ethernet works better in practice than in theory.” In general, Ethernet and token ring had similar numbers of problems, but where Ethernet tended to have performance problems, token ring tended to outright break, and it cost 3-5 times more to buy in the first place.

As for all the “fixes” Tolly details for Ethernet, such as switching and QoS additions, let’s not forget that the original IBM token ring ran at 4Mbps, then was upgraded to 16Mbps, and was supposed to morph into 100VGAnyLan, on Tolly’s list of flops. Switching was also added (and needed), though the price per port was near astronomical. Was Tolly looking back at token ring through rose-colored glasses?

I’d also have to wonder how Arcnet got on Tolly’s list of flops, since it had a life about as long as token ring and wide implementation in small to midsized companies, while token ring was seldom found outside of large companies. Anyway, it’s all water over the dam, and it’s amusing to think of all the fun we had back when there were lots more choices of technology and no one was sure what would win.

Karl Compton

CTO

Cierra Business Solutions

Houston, Texas

I was surprised to see no mention of Ethernet jumbo frames in Kevin Tolly’s column, “Ethernet’s secret source.” I see it is hinted at with the mention of framing sizes and the IEEE taking on extended frames, but I’ve seen people use jumbo frames for years and thought it would have merited mention in this context.

Scott Hofer

Broomfield, Colo.

Arcnet not a flop

In his column, “Ethernet’s secret source”, Kevin Tolly’s calls Arcnet a “LAN technology flop.” I agree that Arcnet did not have a long life, but one should consider what it did for technology.

Arcnet was designed to run on RG-62. This coax was installed to support the IBM 3278 and 3279 terminals of the day that connected to the mainframe. Arcnet allowed the use of PCs on the existing cable plant, something that didn’t happen again until Ethernet attempted to use Category 3 cable, which was designed for voice communications.

Although the throughput and the architecture of Arcnet eventually broke down, it had its place in the history of technology. Probably the one thing that killed it quicker than anything else was its 508-byte packet limit. To send 512 bytes of data, a standard IP packet at the time, required two Arcnet packets.

Dewi Sant

Senior technology consultant

TekSystems

Southfield, Mich.

In his column, “Ethernet’s secret source,” Kevin Tolly contends that today’s successfully proliferating and evolving Ethernet technology is actually, underneath, the IBM Token Ring, which may be “gone, but it was never a flop.” Tolly goes back to the old saw “determinism,” accusing Ethernet of not having it. While stating correctly there has not been a reported Ethernet collision in years, he fails to notice, as so many token ring customers did, how much non-deterministic hell breaks loose when tokens get lost. He reports that Ethernet still does not have large frames, completely overlooking that token ring’s failure might have had more to do with it being too complex, too slow, twice as expensive and, thanks to the old IBM monopoly’s bad attitude, non-standard.

Kevin, I know from experience how painful it is for columnists to admit they were wrong. But you should just admit that I was right about Ethernet, and you were wrong about token ring. But the trick is to admit you are wrong while people still care, which they don’t anymore about token ring (RIP).

Bob Metcalfe

Ethernet inventor

Boston

Digital vs. real life

I enjoy reading Mark Gibbs’ cut on things, which are often thought provoking. As an engineer, however, in Gibbs’ column “Digital lifestyle, Part II”, I read James Bandinelli’s comment about the real world vs. digital somewhat differently. I think the digital life’s relationship to the power grid and sewers is different from horses’ relationship to autos. The auto was invented to replace the horse.

The digital life does not replace the power grid – it requires it, as we have seen too well during the recent Northeast blackout, Gulf Coast hurricane and attempts to normalize Iraq. To me, the point would be that more people are excited about going digital than about seeing that there is power to make it work. Some folks are now working hard to make the power grid better, but relatively few are really concerned about replacing it with a whole new technology that doesn’t depend on oil/gas/coal and miles of towers and power lines so the digital life which depends on power can function during disasters and other hard times. (Even fewer are concerned about problems with the sewers or the fact that the water table is dropping all over the West, but that is harder to relate directly to digital life.)

George Broomell

Kalamazoo, Mich.

Bugged by briefs

In the news brief, “Explorer bugs abound”, we learn that a single bug that can crash Internet Explorer has been found in a beta release. In the next brief, “Firefox fixes a few bugs,” we read of a “highly critical” release of eight Firefox fixes (including some that would allow the takeover of unpatched PCs) and an unpatched bug in Thunderbird.

The tone of the Firefox piece is right: here’s the risk, here are the steps users should take to protect themselves. If the IE piece had been written in the same tone, it would have boiled down to: “Tom Ferris has published a vulnerability in the latest beta release of IE. The vulnerability can be exploited to cause IE to crash. To avoid any associated problems, users are advised not to use beta software.”

It seems to me poor journalism to cast a single bug in a beta as proof that IE is rife with risk, while saying the Firefox risk is mitigated because no one knows of any exploits taking advantage of the vulnerabilities yet. And all the bug does is crash IE? On my computer, IE crashes all the time without a “specially crafted HTML file.”

Dan Riordan

President

On-Tech Consulting

Red Bank, N.J.