Ethernet is not a protocol

Posted by & filed under Uncategorized.

Carl’s pet peeve number three*: Ethernet is not a protocol.

IMS Research found that the number one industrial Ethernet protocol is “Ethernet TCP/IP.” [“Double Your Pleasure, Double Your PROFINET”]

VDC found the leading industrial Ethernet protocol is “Ethernet.”

Ethernet is not a protocol!  I ranted about this before, way back in 2006: “Why Use Industrial Ethernet?”  So today I’ve counted to ten and will move from rant mode to education mode.

Networking is characterized in the ISO/OSI reference model.  Its seven layers define the functions of various aspects of networks. Wikipedia has a complete explanation: OSI Model.  The Internet Model collapses the layers to four.

Layers 1 and 2 are defined by IEEE802.3 which we know as “Ethernet.”  To overstate this: Ethernet is just the Physical layer and the Data Link layer.  By itself Ethernet does nothing; it’s just the “pipe.”  What comes down the pipe?  Whatever IP sends down the pipe (when TCP/IP is used).  IP is at layer 3.  IP sends the message it received up to layer 4, TCP.  So what do TCP and IP do?  Just send the message on to where they’re told.  Up at the top of the stack is the application layer, at layer 7.  The application layer is where the protocol resides.  Only the application layer actually does more than send the message along its way.  (What happened to layers 5 and 6?  They are neither needed nor used.)

Now, IP and TCP are not the only choices for layers 3 and 4, respectively.  You can substitute UDP at layer 4 for example.  Or you can skip them altogether like Address Resolution Protocol (ARP) does; it goes straight from layer 2 to an application.

PROFINET takes a lesson from ARP.  It skips TCP and IP for real-time messages.  TCP and IP (aka “the stack”) is implemented in software.  Rather than let PROFINET real-time messages languish in the stack, we skip it.  This improves speed and determinism.  (See the Industrial Ethernet Book article for details: “Technical Article: Performance metrics for Industrial Ethernet”.)

So, when users respond to IMS and VDC and say their Industrial Ethernet protocol is “Ethernet,” what do they mean?  I would not accept this answer if I were a market research company!  So what do they mean?  Do they mean web server? A real Industrial Ethernet protocol?  A Proprietary protocol?  “I have no blooming idea what a protocol is?”

What do you think they mean?

–Carl Henning

*Number one: Wireless is not monolithic: “Wireless or Wireless
Number two: Why can’t Control Engineers and IT get along?  (Although there is certainly improvement in this relationship from five years ago.)

17 Responses to “Ethernet is not a protocol”

  1. chenning

    Dave, not that I could be biased in any way, but that can’t be right. :) Or at least not as an explanation for all of the answers. I’m guessing that it’s a mix of many “Ethernets.” But what are they?

  2. chenning

    Good point, Victor, on the coexistence idea. They may have a web server connection, an HMI connection, and a connection for their programming software – all on the same wire. Maybe I’m too focused on IO but there are really many other Ethernet-based functions happening out there.

  3. James Ingraham

    (a) I think the “I have no idea” answer is most accurate. I have been railing on this point for a decade, and I *STILL* can’t get it through to most of my company. I’d think they want to learn just to shut me up about it. So unless the surveyors got to actual controls guys and not mechanical / sales / techs / finance / management / whatever, I have NO faith that they would get the right answer. I’ve even had some controls guys not seem to get it, particularly when they only use one brand all the time.
    (b) Why can’t IT and controls get along? Because NOBODY gets along. Much as I despite IT acting like they are the purpose of a business rather than an enabler of it, I can’t really say that they’re any worse than anyone else. Everyone likes to rule over their own little fiefdom. Nobody wants to do extra work. So it’s not those two groups who aren’t getting along, it’s just an obvious friction point. Ask marketing about finance some time; you’ll get a similar reaction.

  4. Rowan Finch

    1. My experience for so long now is that anyone that says Ethernet is an industrial protocol , just has no idea on the subject. Ethernet/IP is also easily confused with those with a fringe knowledge, simply because of the close alphabetic relationship to TCP/IP. Drop the TCP, stick in Ethernet and you have Ethernet/IP as a string of characters at least. For the typical Ethernet response, I usually respond with Ethernet is the Pipe, what is the language that is spoken down the pipe? (taking away all the layer details and linking it to something most understand sometimes aids a better understanding).
    I also find that the term Ethernet as a protocol can be used depending on the control system that person is most familiar with. i.e: RA-Ethernet/IP, Siemens-Profinet. Schneider – ModbusTCP, B&R-Powerlink, Beckhoff-EtherCAT and the list keep getting larger…
    2. IT& Controls headbutting is simply as they are standing on each others turf and it is a natural defence mechanism. Both side in defence typically do not want to understand the demands and issues on each side of the fence. It really come down to a environment where these potential conflicts are eliminated so that productive mutual cooperation can take place.

  5. Fritz Woller

    My take on “What do you think they mean?” is:
    The majority of most likely mean Rockwell’s EtherNet/IP.
    Reason for this assertion is twofold:
    1.) Most of them will reside in the Rockwell dominated market (e.g. North America)
    2.) They have fallen prey to Rockwell’s clever marketing and naming of their Industrial Ethernet protocol.

    Don’t take my word for it though, the following quote is straight from the horse’s mouth (see the January 2012 issue of “Automation Today Asia Pacific” magazin @
    —— Quote ——-
    “Also, EtherNet/IP is the only industrial network built on standard TCP/IP technology – the same Ethernet standard as e-mail, the Internet and other commercial applications.
    Other industrial Ethernet protocols require proprietary chipsets and hardware, which can hinder innovation and system longevity.”
    —— Unquote ——-

    Do I need to say more?

  6. Ian Verhappen

    Wholehearted agreement that Ethernet is NOT a protocol. Ethernet is not much different than say RS-485. I use the analogy that Ethernet is the express service for sending packages while serial is dial-up modem or truck transport. The Sea Container or pallet (the packet) is oblivious to how it is being carried.

    Unfortunately, several of our friends in the trade press and elsewhere inadvertently perpetuate this myth by asking in surveys which fieldbus are do you presently or plan to use, and then include Ethernet as an option. Sort of like asking what is your favorite type of apple and do you like oranges.

    Good news is that this means job security for those that realize it is the protocol that makes all the difference.

    P.S. Ethernet/IP was a GREAT marketing coup by the folks at ODVA

  7. Enrico Querci

    People often have no idea about what a protocol is …

    In the past, when serial protocols still dominate, the classical answer to the “which protocol do you use in your hardware ?” was “an RS-232 protocol” or “a serial protocol” …

    It was like you ask a person “in which language would we speak ?”, and its answer was “by telephone” …

  8. chenning

    James, I also think a lot of it is “I have no idea.” But more realistically it is probably an all of the above.

    Rowan, I like the point about Ethernet meaning different things to different users depending on their vendor of choice.

    James and Rowan, it’s time to revisit the IT and controls collaboration in a new post!

    Fritz, I really think it’s all of the above, not so much Ethernet/IP by itself. It’s funny that Rockwell is stuck on the “EtherNet/IP is the only industrial network built on standard TCP/IP technology.” I guess they forgot Modbus TCP… oh, and PROFINET (which despite their delusion, does not “require proprietary chipsets and hardware). I’ve beat this topic to death like at and

    Ian, I’m latching on to my job security. Ethernet/IP – marketing coup or market confusion?

    Enrico, I myself have said “RS232 is not a protocol.” Blogged about it, too:

    There were a few comments in the PROFINET LinkedIn group, too. Main point there is that we have a lot of education yet to do. Yes, we do.

  9. James Powell

    Hi Carl,
    Not to totally disagree with you, but the word ‘protocol’ has many definitions. One of them listed in the Merriam-Webster on-line dictionary (, states that a ‘protocol is a set of conventions governing the treatment and especially the formatting of data in an electronic communications system ’. Ethernet refers to the physical layer that many different communications networks can run on. Since it does define a ‘set of conventions’ for getting the ‘1’s and 0’s’ from point A to point B, I believe that it does fit the formal definition of what a protocol is.
    However, I also agree that it does not go far enough to be considered a network protocol since it is only one part – you have define the data formatting. I think the part that confuses end users is that unlike Profibus DP, Device or Modbus RTU, you can have multiple protocols running on the same wire. When I install a Profibus DP network, only Profibus DP can run on that network. However, Ethernet is confusing since you could have your office network running on it and then have Profinet and Modbus TCP/IP running on it all at the same time (which I have done).

  10. chenning

    I tried to direct commenters from the PROFINET Group on LinkedIn to come to the blog to comment so all the comments would be in one place. Didn’t work. [You can look at all the discussions in the group here: Here is one comment from the PROFINET LinkedIn Group that reinforces James Powells’ point:

    Rob Hulsebos I beg to differ. You probably mean “Ethernet is not an application protocol” as mentioned as example in the topic start.

    On its own, its a simple (OSI-layer 2) protocol. It has defined the electrical signaling, a message format, it has an address format, detection of transmission errors, retries (sometimes), flow control, point-to-point and broadcast capabilities, medium access control, VLAN’s, QoS, autonegotiation, etc. That you need a little bit more in order to use it as an application protocol, OK. But I’d still give it the description “protocol”.

    A comparison with RS232 is not valid, as RS232 doesn’t even do those things I mentioned above. Therefore it is an “interface”. Even more primitive than Ethernet.

    Are TCP and IP protocols? Which aspects of these make it a protocol?

    There is no strict definition of what a “protocol” is. Hence, this could become a long discussion…

    Rob and James, I’ve enjoyed this tangent as it refines definitions, but the real point I was trying to make and the answers I was looking for got lost. James, it’s always wise to consult the dictionary; I did after I read Rob’s comment, but before I read yours: “computing – the set form in which data must be presented for handling by a particular computer configuration.” (World English Dictionary). Wikipedia comes through with a too-long article “Communications Protocol (” It talks about layered protocols and considers each of the seven layers of the ISO/OSI reference model a protocol. So let’s accept Rob’s adjective and say “Ethernet is not an application protocol.”

    My (not-so) clever headline was to call attention to the fact that Ethernet by itself does nothing more than move frames from point A to point B. It takes an application protocol to do something useful with them.

    >>> Now back to my original puzzle: what does a user mean when he answers that his industrial protocol is “Ethernet.”

  11. chenning

    Ethernet is facial tissue; PROFINET is Kleenex.

    In the PROFINET Group on LinkedIn Paolo Scarfi made what may be the defining observation:
    “Normally the users claim to have ethernet as industrial protocol when they have no matter what messaging on ethernet (modbusTCP, profinet, ethercat, powerlink, sercos, etc.). This is an effect of the commercial/marketing people who never explain clearly what they are selling using the word ‘ethernet’”

    So all PROFINET is Ethernet, but not all Ethernet is PROFINET. Just like all Kleenex is facial tissue, but not all facial tissue is Kleenex. You can continue with Xerox and copy machines, or if you’re a fan of The Big Bang Theory (the TV show not the actual theory), use hot tubs and Jacuzzis.

    I still vote for all-of-the-above as an explanation for “Ethernet is an application protocol,” but Paolo’s observation may account for a lot of “the above.”

  12. chenning

    The award for the “Best Analogy” goes to Paul Chapple on the LinkedIn discussion page: “I often use the analogy of a telephone. I can call any number in China, but unless the other person can speak English I cannot communicate. So we have a compatible system, where data (in this case voice) can be passed, but then each system must ‘speak the same language’ to actually achieve anything.”

  13. Rick Rockershousen

    I agree that Ethernet is not a protocol by itself. If you look at it as part of a protocol stack using the ISO OSI model, then Ethernet is in the stack, albeit at the lower layers. There is definitely confusion over TCP/IP over Ethernet and EtherNet/IP – people who don’t know about Industrial protocols sometimes think that EtherNet/IP is just standard TCP/IP over Ethernet. So ODVA/Rockwell did a horrible job in coming up with the name EtherNet/IP – it is too close in name to what most people think of when they say Ethernet (and mean TCP/IP over Ethernet).

    That being said, when someone calls Grid Connect and says that they have Ethernet in their automation installation, they often say that because they know there are Ethernet cables in their application. They often have no clue if it is PROFINET, Modbus TCP, Allen-Bradley Ethernet, or EtherNet/IP or something else. Usually, after some friendly discussion, we can help them figure it out.

  14. John Morse

    Oh this old chestnut! … I am not aware of any official definition of a protocol in the context of industrial comms; I guess you might know otherwise Carl. However, a typical dictionary definition of a protocol goes something like this: a modus operandi; a code of behaviour; a set of rules; a method of communication; to name but a few. So …. PROFIBUS is one set of rules, Ethernet /IP is another set of rules, Ethernet TCP/IP another set of rules and PROFINET, yet another set of rules. So, in my book they are all protocols, albeit different protocols but all protocols. I rest my case.

  15. systech

    Thank the lord that there are other people out there who have this issue! I constantly seem to rable on at programmers and developers about their terminology! Ethernet has never been an individual protocol but they always refer to it as a ‘protocol’ that they work with.

« Back to PROFIblog