Slashdot Mirror
30 Years of Ethernet
Babylon Rocker writes "An interview with one of the inventors of Ethernet." Metcalfe talks about the history of Ethernet as well as what he's been up to for the last couple years. (Not surprisingly, he's now a VC ;)
← Back to Stories (view on slashdot.org)
w00t!!, now for the serious side.. Ethernet is awesome, but I'm already tired of it. I can't wait to sit back and see what happens with 802.11(alphanumeric character) 30 years down the line.
"Victory means exit strategy, and it's important for the President to explain to us what the exit strategy is." G.W.Bush
What's VC stand for?
I read the article and still don't quite get it
Does anyone have BitTorrent for the past 30 years of the internet? I really need it. Thanks!
Karma: The shiznight, mostly because I am the Drizzle.
Happy Birthday Ethernet!! Hurray for Ethernet! May it live long and prosper, and my bandwidth never end.
Metcalfe has a habit of saying stupid things, I wonder why people keep listening to him. One great invention thirty years ago, paired with a huge ego, does not an oracle make.
314-15-9265
Wait... if Ethernet has been around 30 years, that makes TCP/IP PRETTY DAMN OLD!!!!! Anyone up for re-inventing the wheel??? Maybe someone can make a protocol in which practically any piece of information can be traded, with a special way to commit a special pipeline to different medias (such as movie/music downloads getting a compressed, special set of ports used just for that purpose..).
Next thing you know, the teleco's will be bringing up charges against us for inventing a better internet... when will this end!?!
OR OR Or......... Maybe I'm delusional...
"Victory means exit strategy, and it's important for the President to explain to us what the exit strategy is." G.W.Bush
VP=venture capitalist
Wow, it's been 30 years already, and i *still* haven't managed to get my mitts on a set of RJ45 crimpers...
0110100100100000011000010110110100100000011000100
An interesting but old article on wired about Metcalfe here: The Legend of Bob Metcalfe
He joined the Vietcong? That's pretty suprising!
Well they do say vietnam is a beautiful country now that American invaders aren't trying ton bomb into the stone age and kill everyone who doesn't dig bigmacs...
If you're a "free software" true believer, I guess that column must irritate. But I don't see anything stupid about it. Care to get specific?
According to Metcalfe, Ethernet is competing with SONET and Fibre Channel, although he claims that ethernet is winning due to its "internet-compatibility," among other things. To me, this seems like steps in the wrong direction. If fibre optics do not integrate well with the present structure of the internet, then we should be changing the structure, not sticking to the old concept of ethernet. When ethernet was invented, it took advantage of technology available 30 years ago. Since then, we have only been improving on the implementation. Despite the fact that SONET and Fibre Channel are the current "Godzillas," THEY are the ones with the novel technology, and avoiding them would not be in the best interest of advancing technology.
Would you rather be stuck with Token Ring?
I mean, IBM is a great innovator to be sure. But token ring, IMHO, was one of their great misses. Running at 4 or 16 Mbit, it was certainly speedy for its day, but the protocol to implement it is hellish, not to mention the extra hardware to make it work. It's really much more involved than the plug-'n-play 10baseT that we've all gotten used to with the dominance of Ethernet.
I'm sure you've heard the joke--Why are you all on the floor? Someone pulled out the network cable and we're looking for the token. Anyway network topology that relies on peers to propagate a piece of information to its neighbour is doomed to failure.
Karma: Excellent Birds (mostly as a result of listening to Laurie Anderson)
What, no gratuitous Al Gore comments? ;)
Second, Metcalfe defines "broadband" to mean "high bitrate" rather than "uses a broad frequency band". Nitpickers like me have been quibbling over this change in definitions, but if someone like Metcalfe has gone over, it's time to let it drop!
"The technology is 30 years old...who'd want to use it?"
Wasn't that one of Microsoft's arguments against Linux at one time?
And, I *KNEW* I was a geek when this kept me laughing for 30+ minutes...laughing so hard I had tears rolling down my eyes and my sides hurt:
Ethernet: A device used to catch the Ether-bunny.
{snerk...hahahahaha}
Have you read the moderator guidelines? Well, have you, PUNK? (and I want a Karma: Gnarly option)
Okay, not originally said about ethernet... "How can one little insulated wire bring so much happiness??"
/sig
Just raise the taxes on crack.
Anyway network topology that relies on peers to propagate a piece of information to its neighbour is doomed to failure
That's not what the RIAA would have you believe!
in girum imus nocte et consumimur igni
"The lesson to be learned is not to take the comments on slashdot too literally." --Vinnie Falco, BearShare
If you have, you would absolutely, certainly have the initials VC engarined in your mind. These are the overlords that controls your life and owns your soul.
Especially toward the end when all of them were changing from benevolent take-all-you-want piggybanks* to bloodsucking vampires that fires off one coworkers after next with glee**.
*note1: actually, from the beginning it was more like the inverse of beggars: they often *BEGGED* you to take their money if you just had the stupidest business plan involving the word "internet" and "e-commerce."
**note2: okay, I have to admit they didn't want to see the company they have vested interest fail, but toward the end, most VCs took control of their companies directly, and had no quarrals about tossing people out like used rags.
For all the geeks out there - the whole dot-com -> dot-bomb thing taught me one big lesson: unless you make it to upper management or start out on your own (really on your own, i.e. your own capital), you are just a (disposable) pawn in this game.
My life in the land of the rising sun.
Sorry, what extra hardware are you referring to?
It's really much more involved than the plug-'n-play 10baseT that we've all gotten used to with the dominance of Ethernet.
10BaseT also required extra hardware, in the form of a hub or switch. Unswitched ethernet is also susceptible to many similar problems, such as a single broken or malicious node that won't stop transmitting. Older readers may remember how expensive ethernet switches were, which is why some sites were forced to deal with the nightmare that is 10Base2.
Before 1993 or so and the advent of Switched Ethernet, Ethernet would melt down under the weight of its own traffic. 40% traffic for Ethernet is an emergency situation. I've seen TR networks hum along with 80-90% utilization and the users barely know.
Token Ring has built-in QoS. It has several levels of error monitors. These are things that are kind of added by switches, but are not a fundamental part of the topology. And if you don't have a *good* switch, you don't even have that.
Of course, in the early 90's Ethernet cards were under $100 and Token Ring cards were $400. *That's* why Ethernet won. Not speed: TR was doing 16MBit when Ethernet could only do 10, and remember, I can acually *get* 16Mbit from TR, instead of 4Mbit with Ethernet. Today, with good switches, I don't miss TR too much. But before switches...
Maybe that's why many, many very large organizations were using TR even into the early '90's. Try running 300 computers on unswiched 10Mbit Ethernet (the best Ethernet had then)...
Linux IT Consulting and Domino Development in Michigan
"Would you rather be stuck with Token Ring?"
Ya, just look at what happened to Frodo!
"Much work is lost, for the lack of a little more." -Edward H. Harriman
Admittedly, as Token Ring was dying, people were finally making the Token Ring equivalent of an Ethernet switch, with each port a ring unto itself, but by then it was way too late.
As for the speeds, I've only ever heard of 4 and 16 Mbps. I'm sure that there are optical protocols that pass tokens, but they're not Token Ring [tm].
--
"Open source is good." - Steve Jobs
"Open source is evil." - Microsoft
It's not a choice of photons vs electrons. Ethernet can use optical fibre too.
And I'm glad it's not up to you which technology we use, because the actual tech is only one part of the overall usefulness of a technology. For example, a $100 network card that can do 1gbit/s is more useful to me than a $1000 network card that can do 100 gbit/s. Because I can afford (and justify) the $100 card.
Price matters. Open standards matter. Would we have Ethereal and tcpdump and all the billions of useful network tools that are out there if we were using proprietary standards for networking? I don't think so. Would people get owned due to network stack (or network protocol design) bugs? Seems quite possible.
Try looking a little farther out.
Token passing in TR was pretty basic. The token travelled the electrical ring and each node passed it through. If the node had a packet waiting for transmit, it listened for the bit (actually a pair of Manchester encoding violations, IIRC) that indicated the current packet was a token and converted the violations into valid bits, essentially stripping the token and turning it into a data frame. Now there were other complexities like having a master station and error recovery, but the basics were pretty simple.
I wonder what would be the world without Ethernet? Would Internet begin and survive just on UUCP/SLIP/PPP?
Would you rather be stuck with Token Ring?
I mean, IBM is a great innovator to be sure. But token ring, IMHO, was one of their great misses.
There's token rings and there's token rings. Saying "Token Ring" when you mean "IBM's Token Ring" is like saying "DOS" when you mean "Microsoft's Disk Operating System".
My guess is that, in the absense of the invention of Ethernet's listen/transmit/back off on collision model, we'd have bootstrapped up from Datapoint's ARCnet.
Like IBM's protocol, ARCnet is also a token ring. But unlike IBM, ARCnet's transport layer is broadcast. So it combines the self-healing characteristics of Ethernet with the delivery-time and latency guarantees of token rings.
ARCnet did have a downside - limited number of addresses on a segment. But there are ways around that. (My favorite is my own variant which I call "bumblenet", involving an aborted binary search for the next station to get the token.)
But for radio, Ethernet-like low-level protocols have a distinct advantage over token rings: They suffer less from the "hidden transmitter" issue - where some devices can hear each other and others can't.
Ethernet-like protocols get their packet through if the transmitter and receiver can hear each other and nobody else within their earshot is talking. So a pair of stations on THIS side of the hill can swap a packet at the same time a pair of stations on the OTHER side are swapping one, without explicitly negotiating about it over relays through other stations - or even knowning about stations on the other side of the hill.
Token rings can work around a hill - even if you have a sparse chain of stations where each can only hear two neighbors. But there's a lot more effort involved. They only get simple when either everybody hears everybody else or everybody talks to exactly two neighbors in a closed ring.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Whopptie-doo.
Beta was better than VHS, MicroChannel was better than ISA.
The problem with token ring was that it was a proprietary technology hocked by IBM. If you aren't a large government or corporate entity, IBM doesn't give a shit about you.
Conformity is the jailer of freedom and enemy of growth. -JFK
Bob Metcalfe once predicted that the internet would 'gigalapse' due to IP namespace exhaustion and sheer load. It didn't happen.
Bob has made a career out of making an ass of himself with idiotic predictions coupled with a humongous ego. He fancied himself quite clever when he called the free software/open source movement "the open sores movement." Har har! You may have a career with ZDNET yet, Bob.
Hey Bob we thank you for ethernet, but you're still a jerk.
is CSMA/CD . What a brilliant MAC. You just start shouting, check to see if anyone else was shouting, and if they were, wait a random amount of time and start shouting again. It's so simple and stupid that no one would ever think it works.
simon
home page
He makes this rather ignorant comment:
Open source contributors who use the GPL never "give their intellectual property away". Copyrights are very strongly defended; the recent FSF vs OpenTV story is sufficient proof of this. Trademarks are very strongly defended: Linus and RedHat have both defended trademarks. Patents are a sticky mess but even then the GPL doesn't demand that you give up patents; only that you don't use them to restrict or impede licensing. The open source movement is not so stupid as to "give away" code. Strong ownership of intellectual property is at the very core of open source.
The subtle but important distinction is that open source developers want to share their intellectual property. The philosophy is "you may use my IP if I can use yours". This is not giving anything away; it's building a community of cooperation. There is an exchange of value between two parties even though the exchange is not monetary.
I suppose it's possible to argue that BSD zealots are giving their intellectual property away. Yet another reason to avoid the BSD license.
...since I had to fsck with terminating resistors (10base2), vampire taps (10base5) or that crappy stiff ArcNet cable. I still have working setups of each packed away in my basement.
Everyone will start to cheer when you put on your sailin' shoes.
No, you're kidding me, right?
I can't believe this FUD is still out there after 30 years. Contrary to popular and mis-guided belief, an Ethernet will NOT saturate itself at 37% utilization. Period. Anyone that honestly believes that should give the token ring and ATM salesdroids and spin doctors a great big pat on the back because that's exactly what it is: sales doubletalk and spin from vendors of competing technologies. For christ's sake, this myth was laid to rest in September of 1988 . This FUD relies on over-simplifications of assumptions in the theory and inadequacies in the testing procedures.
I can't believe you'd honestly bring it up. Anyone with even a marginal amount of actual networking experience knows this to be FUD. Next time think before you speak about something you know nothing about.
" Linux is based on thirty year old technology, and it hasn't replaced Windows. He's right. You know it."
Windows is based on 30-year old technology and hasn't replaced Unix. In fact, the dominant form of computing 30 years ago was the mainframe. Windows hasn't even replaced that.
It's fun to to go down memory lane sometimes - remember ....
.... you know when one old piece of shit Broken Ring NIC that decides to take a header- could bring a Fortune 500 company to its knees ....
"OMFG! The Network is Beaconing! The Network is Beaconing!"
Ahhh the good old days
Frankly - ArcNet was better!
---- "Logoff! That cookie shit makes me nervous!" - A. Soprano
Actually, TCP doesn't expect 0% packet loss at all. It expects an unreliable data path.
The problem here, as you subsequently covered, is that ethernet backs off and handles its own retransmission. This screws with the TCP timers.
If you could turn off the ethernet retransmission then TCP would still work just fine.
But yes, unswitched ethernet sucked when you had many stations trying to move large amounts of data. Token ring (and ATM!) _did_ handle load better but was much more costly.
(My DEC friends commonly refer to Ethernet as "cheapernet" with a little curled tongue action.. eww.)
Linux sucks because it's 30-year old technology even though it's first public release was in 1991 and it doesn't have any ties to any of the heriditary unicies. (Making an assumption about the SCO claim, of course). Well, gee, brilliant. Knock down something simply because it's old. Let me tell you something folks, stop using wheels, old technology, innovate. Cars? 100 year old technology, innovate, come up with something radicly new! Let's face it, either this guy is sleeping with aliens, or he's a hypocrite. Go him!
-If God wanted people to be better than me, he would have made them that way.
Boggs invented the first (of many) hardware circuit techniques to do collision detection, and other elements of transceiver design. If Dave hadn't picked up a soldering iron, we'd probably be doing DATAKIT or some other telco hack.
...-.-
See:
Measured Capacity of an Ethernet: Myths and Reality, David R. Boggs, Jeffrey C. Mogul, Christopher A. Kent .
Ethernet is CSMA/CD, not CSMA. The collision detection mechanism arbitrates access to the medium, it is not there for flow control. Collisions are not bad.
Mea navis aericumbens anguillis abundat
Oh hey, found a Commmunications of the ACM article by Metcalf and Boggs about their early work on Ethernet. Good read.
I wonder how long we can keep this up?
A Pentium processor is 10 somewhere around these days (don't remember exactly ;)
You really don't believe the entire internet runs on one giant ethernet network do you?
This is why I hate being a network junky, anyone who owns a ethernet hub and some cable and plays counterstrike over their home network believes they are a networking genius because they know all there is to know about the one and only networking protocal.
So no, don't thank him to much, without ethernet, we'd just use something else for cheap, efficient LAN's, vaccums tend to be filled.
--Nuintari
slashdot : where an opinion can be wrong.
This way carriers will soon be offering ethernet access, without having to abandon their ample investments in SONET.
Expect to hear about this technology a lot in the coming years.
Read more here.
Enter EoS in the searchbox at the upper left.
Flourescent (adj): smelling like ground wheat.
Kind of like the Jobs-Wozniak team. One gets the glory while the other languishes in realitive obscurity. Were's the Wozniak reality-distortion field?
It's that smokey haze that hovers around him...
Issues of load-based meltdown aside, token ring has one property that ethernet does not - token ring is deterministic with regard to delivery time. When you're writing B-52H simulator code, you _really_ want your frames to get there on time, every time. (where a few milliseconds late means glitchy flight).
Tiller's Rule: Never use a word in written form that you've only heard and never read. You will end up looking foolish.
TCP expects a network with 0 packet loss
I can't believe you just said that. Um, the whole reason that TCP exists is to provide reliable services on an unreliable network (that is, > 0 packet loss).
I also can't believe they are letting idiots like YOU into MIT now...
God, I love this world.
Bob may well hold passionate and horribly misplaced beliefs, but he is open to arguement and, on occassion, can even change his mind. He also was/is the driving force behind Pop!Tech, an annual brain candy event where the questions around "what does all this tech mean" are addressed. Most people seem to miss that Bob is a provocateur...he *enjoys* annoying the crap out of people and forcing them to defend their often dogmatically (and poorly understood) positions. Mind you, this is not to say that Bob does not also hold some opinions dogmatically...but at least he can defend *why* he hold them. Personally, I like a man (or woman) who takes such pleasure in forcing others to think and defend their positions...in trying to respond to his taunts, deepen the understanding of their own positions, is this a bad thing?. Then again, maybe it is easier just to say that he's a bigmouth...perhaps you can force him to drink the hemlock.
rootrot
--
The reasonable man adapts himself to the world: the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man.
-George Bernard Shaw
My home office is going to remove the token ring network next month.
Lasers Controlled Games!
They've even tried migrating to Ethernet. They will switch a division from time to time, but they've found that Ethernet actually costs them more: they have to buy new Ethernet switches compared to the cost of continuing with TR (which for them is near zero, with an inventory of 100Mb cards and CAU's already purchased), and it buys them pretty much nothing. So until there's an actual reason to, they're staying TR...
Linux IT Consulting and Domino Development in Michigan
It would be: Try to drive to your destination, and if you hit someone, go home and try to drive there again. (-:
Someone named an OS for me.
Thanks for clearing that up. I *think* what I meant to say is that if there's only one route to the gateway, then TCP gets fubared if the data path is unreliable. Of course, this would be the case with any protocol - what TCP was designed to fix is to choose an alternative path given that they exist. With Ethernet and TCP, though, when you have a heavily used link, it's particularly bad because of the retransmission delays you pointed out.
Eep!
TCP doesn't care about alternate data paths. Thats an IP routing issue.
You could, in theory, implement TCP directly on top of some other protocol quite happily.. including directly on top of ethernet.
Its all in the timers..
"And the reason that Ethernet won was that Ethernet was an open standard with many competing companies providing it."
Now if only more compnaies would take heed...
And who tends to promote interoperability more than Free Software and Open Source authors?
Some people don't care about the difference, but Bob demonstrated his ignorance by stating that RMS is an "open-source software guru," when Richard doesn't even agree with Open Source. Bob failed to notice that there are idealogical, business, and pleasure (Linus did it just for fun) reasons for Free/Open Source software. In general, Bob's attitude seems to be that the bottom line is all that matters, which is silly, IMHO.
You're not delusional as much as ignorant. First, IPv6 has existed for a number of years. It is not a reinvention, but a an evolution of IP to make it more scalable. Second, the value of IP and the Internet is that they are generic, not burdened by application specific details. There's a reason for the protocol stack: to keep application details at the top.
I didn't know what VC meant until I read the article (which I did before reading comments). Of course, maybe /. posts should be written in such a way that those who haven't read the article look dumb. That would make it easier to weed out a lot of clueless posts.