Are always three copies of every packet sent?

[a1111]

Hello,

What do you think cisco meant by the following?

"
Packet switching is a method of data transmission in which a message is broken into several parts, called packets, that are sent independently, in triplicate, over whatever route is optimum for each packet, and reassembled at the destination. Each packet contains a piece part, called the payload, and an identifying header that includes destination and reassembly information. The packets are sent in triplicate to check for packet corruption. Every packet is verified in a process that compares and confirms that at least two copies match. When verification fails, a request is made for the packet to be re-sent.
"

Source:
https://www.cisco.com/c/en/us/products/switches/what-is-a-wan-wide-area-network.html#~what-it-is

This is the first time I've read that three copies of the same packet are sent, for each and every packet.

Thanks.

[MHM Cisco World]

For each and every ping packet not all packet.

That why you get multi reply for each ping.

MHM

[a1111]

Yes, traceroute by default sends three probes, and ping also sends several, but this article seems to be talking about all packet switched traffic. "Each packet contains a piece part, called the payload, and an identifying header that includes destination and reassembly information. The packets are sent in triplicate to check for packet corruption. "

[MHM Cisco World]

The l3 device dont duplicate any packet' it not originate this packet so it only add header and forward.

Some server do duplicate for secuirty' but not network device(like router and SW)

MHM

[Ramblin Tech]

I hit the "Feedback" link at the bottom of the referenced page and asked for someone to review the provided definition of Packet Switching, as it is obviously in error. Perhaps AI generated?

[Giuseppe Larosa]

Hello @Ramblin Tech ,

I think they took some specific feature that applied may be to X.25 but not to packet switching in general I agree that definition is wrong.

Frame Relay or ATM are examples of  WAN frame switching and WAN cell switching with ATM using AAL adaption layers but they do not triplicate each frame or cell they send.

Hope to help

Giuseppe

[Ramblin Tech]

Hi @Giuseppe Larosa 

I tried mentally giving the definition author (machine or human) the benefit of the doubt as they said "a message is broken into several parts, called packets". As you mentioned, this is true of ATM's slice-and-dice and even TCP's MSS being broken into MTU-size chunks. But when it was followed up with "The packets are sent in triplicate to check for packet corruption. Every packet is verified in a process that compares and confirms that at least two copies match.", well that was a bridge (or maybe router) too far for me. I believe that the triplicate-and-compare is what is really catching everyone's eye and needs to be edited out of the definition.

[Joseph W. Doherty]

Indeed, sending three packets is the most curious part.

It would assume a rather high BER, otherwise just send two packets, and compare.  If they don't match, then request another packet to match against the two already received.  If still no match, continue requesting additional packets until two match or give up after so many additional packet requests.

Also, in the original, if no match of two for the original three, only a single additional packet is requested, not another three?  If the latter, do we again only compare beween the second set of three, or between all six?  If the latter, what if we have multiple pair matches, but those pairs don't match each other?  Hmm, getting complex, eh?

Again, traditional, WAN bandwidth was expensive and limited.  (Proof?  Classful addressing to save four bytes.)  Another huge advantage of a packet having its own embedded data verification, each hop can verify whether the packet was corrupted and not forward it.  Further, with something like traceroute, we could, from the source, identify the hop causing high data corruption.

This approach is so bad, I would be surprised if it was ever used, except perhaps as an experiment (laugh, perhaps to demonstrate how bad it is).

Although the triple packet send is itself curious, usually, I believe, packet switching concerns itself, principally, with getting packets from source to destination, not so much with guaranteed delivery or sequencing.  I.e. the proverbially "best effort" to just get packets from source to destination.

However Cisco created this material, it apparently wasn't vetted for quality/correctness either.

[Joseph W. Doherty]

Without some additional context, I think it's nonsense.

[Joseph W. Doherty]

BTW, as your reference appears to be explaining packet switching in a generic fashion, you could do what's described, for data transfer validation, but I'm unaware of this approach ever being used.  Especially in early WAN networks, bandwidth was a very expensive and limited resource, so sending a packet three times would be very high on the priority list of finding a better way.  The better way was including minimal additional data to validate whether data was corrupted in transit, such as parity bits, checksums, CRCs and ECCs.

If you would like a broad overview of packet switching, you might start with: https://en.m.wikipedia.org/wiki/Packet_switching.

If we were discussing IP or OSI packet switching, what's described is far, far different from it.

As an aside, two other approaches for using digital WANs, are circuit and message switching.  Unsure the latter was ever used unless messages are sized bound, and if they are, it's not true message switching.