I have a question regarding the TCPStream package in Rust. I want to read data from a server. The problem is that it is not guaranteed that the data is sent in one TCP package.
And here comes my question: Is the read message capable of reading more than one package, or do I have to call it more than one? Is there any "best practice"?

CodePudding user response：

From the user space TCP packets are not visible and their boundaries don't matter. Instead user space reads only a byte stream and writes only to a byte stream. Packetizing is done at a lower level in a way to be optimal for latency and bandwidth. It might well happen that multiple write from user space end up in the same packet and it might also happen that a single write will result in multiple packets. And the same is true with read: it might get part of a packet, it might get the payload taken from multiple consecutive packets ...

Any packet boundaries from the underlying transport are no longer visible from user space. Thus protocols using TCP must implement their own message semantic on top of the byte stream.

All of this is not specific to Rust, but applies to other programming languages too.