I'm not quite sure what exactly inherent communication is. If artifactual communication covers everything related to how communication actually occurs (on the system level, at least), then how does knowing how communication occurs in ideal circumstances help us understand how communication actually works? Or is my understanding of artifactual communication flawed as well?
This comment was marked helpful 0 times.
yanzhan2
Inherent communication is what the program requires, typically in program and how you write the program, it is minimum communication needed. But artifactual communication happens in the hardware implementation, like the cache line example which the programmer do not expect, so it would be slower if something undesired happens in the hardware.
This comment was marked helpful 2 times.
uhkiv
I think one example of artifactual communication would be retransfer of lost packets in the TCP protocol. The program does not inherently require this retransmission, yet it is needed in practice due to imperfect communication channels. Another example would be the SYN packet when we start TCP, since if communication channel were perfect, we would not need the receiving node to send a SYN to the node who started the connection.
This comment was marked helpful 0 times.
bwasti
@uhkiv
I believe the SYN packet constitutes inherent communication. I believe the line is drawn at the point where it becomes unnecessary to send the data in an idealized world. A node necessarily needs to know that another is communicating with it to initialize the actual transfer of data. Similarly the sending node needs to know that the communication is possible at all.
However, I believe your point brings up the fact that the abstraction level of this definition is not rigorously defined. Is the communication of the size of the data artifactual? I'd argue it isn't, because it could not possibly be known by the processor beforehand. But we are idealizing to have "perfect knowledge about what is needed to communicate," which makes this unclear.
This comment was marked helpful 0 times.
benchoi
I'd argue that sending the size of the data is artifactual communication, since you don't strictly have to send a certain amount of size data to solve the problem. You could send less size data by sending more data per message (and therefore fewer messages).
This comment was marked helpful 0 times.
nrchu
I don't know if I agree with that, ben. Simply because you could send it in a smaller size doesn't mean that it is better to do so. Perhaps each message is independent and it doesn't make sense to group them together, and you would like to know the individual size anyway (such as instant messaging). In general I disagree with the notion that because you can compress the data being sent, it is somehow better. For example, compressing a file might mean that you're sending "less" data but if the latency is low enough you would not want to tar and untar a file every time you send it across.
I'm not quite sure what exactly inherent communication is. If artifactual communication covers everything related to how communication actually occurs (on the system level, at least), then how does knowing how communication occurs in ideal circumstances help us understand how communication actually works? Or is my understanding of artifactual communication flawed as well?
This comment was marked helpful 0 times.
Inherent communication is what the program requires, typically in program and how you write the program, it is minimum communication needed. But artifactual communication happens in the hardware implementation, like the cache line example which the programmer do not expect, so it would be slower if something undesired happens in the hardware.
This comment was marked helpful 2 times.
I think one example of artifactual communication would be retransfer of lost packets in the TCP protocol. The program does not inherently require this retransmission, yet it is needed in practice due to imperfect communication channels. Another example would be the SYN packet when we start TCP, since if communication channel were perfect, we would not need the receiving node to send a SYN to the node who started the connection.
This comment was marked helpful 0 times.
@uhkiv
I believe the SYN packet constitutes inherent communication. I believe the line is drawn at the point where it becomes unnecessary to send the data in an idealized world. A node necessarily needs to know that another is communicating with it to initialize the actual transfer of data. Similarly the sending node needs to know that the communication is possible at all.
However, I believe your point brings up the fact that the abstraction level of this definition is not rigorously defined. Is the communication of the size of the data artifactual? I'd argue it isn't, because it could not possibly be known by the processor beforehand. But we are idealizing to have "perfect knowledge about what is needed to communicate," which makes this unclear.
This comment was marked helpful 0 times.
I'd argue that sending the size of the data is artifactual communication, since you don't strictly have to send a certain amount of size data to solve the problem. You could send less size data by sending more data per message (and therefore fewer messages).
This comment was marked helpful 0 times.
I don't know if I agree with that, ben. Simply because you could send it in a smaller size doesn't mean that it is better to do so. Perhaps each message is independent and it doesn't make sense to group them together, and you would like to know the individual size anyway (such as instant messaging). In general I disagree with the notion that because you can compress the data being sent, it is somehow better. For example, compressing a file might mean that you're sending "less" data but if the latency is low enough you would not want to tar and untar a file every time you send it across.
This comment was marked helpful 0 times.