HTTP is an text based interchange format with a lot of... history. It is currently specified in RFC9110 for general semantics. The next few RFCs define various other aspects (i.e.: HTTP/1.1 is RFC9112).

Semantics

In general, a request-response pair follows the same pattern:

GET /path?query#fragment HTTP/1.1
Field1: abc, def
Field2: xyz
Field1: uvw

HTTP/1.1 200 OK
Field3: abc, def
Field2: xyz
Field3: uvw
Transfer-Encoding: chunked

32
Hello, world!
This is a request.
0

Notably, this used:

• Every line except the response body has an invisible carriage return before the line feed. (They're not actually present in that code block because markdown sanitizes it.)
• Multiple fields (headers) of the same name. For most headers, this has the same semantics as joining all the field values with a comma separator. There are exceptions (Set-Cookie), so it it safest to store the values separately¹.
• HTTP/1.1 chunked coding to indicate framing for the body². This would allow another request to be sent on the same TCP connection the persistence requirements are satisfied³.

HTTP/1.0

HTTP/1.0 has one request and response for every TCP connection.

HTTP/1.1

HTTP/1.1 can have more than one request and response for every TCP connection. The end of a request message is identified by either a Content-Length header, Transfer-Encoding: chunked and a terminating, zero-length chunk, or immediately following the headers⁴.