I’m promoting an anonymous comment to the main blog. This was a comment to How Ethernet/IP Corrupts. I prefer signed comments, but this one raises some good points that deserve continuing discussion:
Let me start out by saying that this is a very interesting post – and it brings to mind a couple of points to be clarified. First, the use of *most* EtherTypes is typically associated with old, legacy and often no longer used or being phased out protocols, such as DECNet or LAT. While this model certainly is clever in this “bypassing the stack” argument, it should be pointed out that by the fact that the stack is bypassed, this model becomes non-routable. This is clearly a design limiting factor in the sense that one cannot design a distributed real-time control environment using this protocol. In addition to “bypassing the stack” this design bypasses every other network – limiting this protocol to Ethernet only. One cannot use other networks such as ATM, SONET, Frame Relay, EV-DO, GPRS or any other network type. Other design challenges are evident in the Ethernet network itself – for example, this design philosophy could lead to very large, flat Ethernet networks, which could produce poor performance. At this point, it uses Ethernet, but not TCP/IP which makes many more things possible. It might as well be another legacy bus network.
The “corrupts the stack” argument is not in the skipping part, it’s in the proprietary hardware discussion. That’s where the perversion of Ethernet and the stack takes place. Your readers should see this site for more details: http://www.industrialethernetu.com/courses/402_4.htm, section entitled PROFInet-IRT. I even like the picture of the corruption– seen here: http://www.industrialethernetu.com/images/courses/Profinet.jpg
Please do not respond to this post with “well, why would you want to,” as this is not a valid argument. The arrogance of most vendors is evident in comments like this. Vendors need to allow their customers the ability to use their products in ways that foster innovation and drive business. This business is full of examples where we use products and services in manners that were never envisioned by their respective creators (i.e. Microsoft never dreamed that DCOM would be used in control type applications).
Thanks for the read and the open forum. Merry Christmas.
The author makes several points that I want to continue talking about:
1. EtherTypes are passé
2. PROFINET is not routable leading to performance issues
3. PROFINET does not use TCP/IP
4. PROFINET will not work on other networks like ATM, etc.
5. “Proprietary” hardware
6. Vendor arrogance
1. EtherTypes are passé
EtherTypes are still widely used for important, every-day functions like ARP.
2. PROFINET is not routable leading to performance issues
3. PROFINET does not use TCP/IP
PROFINET does use TCP/IP – just not for everything. PROFINET uses TCP/IP for configuration and parameterization, for example. PROFINET uses TCP/IP (or, optionally, RT) for peer-to-peer integration. PROFINET IO using PROFINET real-time is not routable because it is a layer 2 protocol. PROFINET CBA (Component Based Automation for peer-to-peer) can use TCP/IP and is therefore routable. Performance is not an issue as a recent independent study confirms.
4. PROFINET will not work on other networks like ATM, etc.
“Well, why would you want to.” [Sorry, I couldn’t resist the temptation.] Seriously, PROFINET is one of several Industrial Ethernets. Ethernet is part of what PROFINET is, by definition.
5. “Proprietary” hardware
Proprietary? Slowly I turned…step by step…inch by inch… [Sorry again. Watched too many Three Stooges shows in my youth. Although the slowly-I-turned bit is a good one and it does describe my reaction to this word as misapplied here and frequently elsewhere, too. Actually the shtick dates back to vaudeville – I wonder if my grandfather and his brother used it when they were on the circuit. If you have no idea what I’m blathering about, check Wikipedia.] First of all, PROFINET RT (Real Time) does not require any special hardware – proprietary or otherwise. It achieves its superior level of determinism by simply bypassing TCP/IP. It is happily transported through standard switches and standard WiFi. The referenced Industrial Ethernet U article is on PROFINET IRT (Isochronous Real Time). Used about 10% of the time and mostly for motion control, PROFINET IRT uses the same bypass through the ISO/OSI layers that RT uses. The magic of PROFINET IRT is in reserving bandwidth on Ethernet. It is entirely independent of the stack, so there is no “corruption of the stack” here either. The referenced image just shows the bandwidth reservation; there is nothing there related to the stack. Now, turning slowly to “proprietary:” there is nothing proprietary about PROFINET! PROFINET is an open standard. Anyone can buy the spec from IEC and implement PROFINET in software or hardware. Several vendors of chips have done so and you can buy that non-proprietary hardware from NEC, Siemens, Hilscher, and shortly, from others.
6. Vendor arrogance
I’m not with a vendor. I’m not arrogant (in my humble opinion), although I am a bit of a smart aleck, perhaps. PTO like PI (the umbrella organization for the regional organizations like PTO) is a non-profit, member-supported organization. Our agenda is simple: to promote the technology and to help members promote their products.
“This business is full of examples where we use products and services in manners that were never envisioned by their respective creators.” True enough. PI continues to respond to and anticipate users’ uses through the extensive Technical Committee and Working Group structure within the organization.