I wanted to know an effiecient way to delimit the data I am sending in network streams while maintaining the smallest possible size of the message.  I am sending back to back messages and need a way to separate them.

Also I have a feeling someone is going to say, use the length/size of message to split them, which I have no problem doing it if is the best and most effiencient way.  And if size is the best way, what is the best size to send excluding headers.  please give answer in bytes.

There are two ways to delimit:

1) size/data
2) implicit length

The implicit length can be used if you know how long each message will be, based on the first byte or two of the message. For example, if you identify message types with a byte, then you can read a single byte, and then dispatch to a handler based on that value. The handler would know how to read the data for the message. Each data type within the message (string, blob, etc) would have to be perfectly delimited or length-prefixed, though, or you wouldn't know where a string ended, for example.

In general, it's easier to deal with data where you already know you've read the right amount, so even if it's less efficient, size/data is often used. To encode integers, especially for size, I often use a "continuation bit" method. I store 7 bits in each byte, and if I haven't stored enough bits yet, I set the high bit, and store another 7 bits in the next byte. Repeat as needed :-) (For signed numbers, you also want to store the sign bit in the second-highest bit of the first byte for negative numbers, but sizes are never signed.)