Rensselaer Polytechnic Institute CSCI-4220 – Network Programming David Goldschmidt, Ph.D.

 A protocol is an agreed-upon convention that defines how communication occurs between two (or more?) endpoints  All endpoints must “understand” and correctly implement the protocol  Protocols must be formally defined, unambiguous, and well-documented  Protocols should address error conditions and unexpected scenarios

 Network APIs provide the bridge between applications and protocol software  Services are made available (often by the OS) Application Network API (via the OS) Protocol X Protocol Y Protocol Z But also OS-independent!

 Network API often provides a generic programming interface:  Support for multiple communication protocol suites/families (e.g. TCP, UDP, IP)  Endpoint address representation independence  Network data types (for portability) ▪ With from-host and to-host conversion functions ▪ e.g. htons(), ntohs(), htonl(), ntohl(), etc. Application Network API (via the OS) Protocol X Protocol Y Protocol Z

 Functions provided by the Network API include:  Specifying communication endpoints  Initiating a connection (e.g. for TCP)  Waiting for incoming connections  Sending and receiving messages  Terminating a connection  Error detection and handling Application Network API (via the OS) Protocol X Protocol Y Protocol Z

 A socket is an endpoint for communication  Communication takes place over a pair of sockets : client server :21202 socket :80 listener socket :9500 socket

 Pitfalls of socket-based communication between client and server include:  Once a server binds to a port, no other server may bind to that port  If client and server do not obey the rules of the protocol, errors may occur  Client/server communication must often be synchronized

 Each process has a file descriptor table (maintained by the operating system)  This table is inherited from the parent process  Defaults to stdin, stdout, and stderr  When open() or socket() is called, the next available descriptor is assigned descriptor (int)default 0stdin 1stdout 2stderr

 Socket descriptors are used to keep track of open socket connections descriptor (int)default 0stdin 1stdout 2stderr 3socket 4 5 etc. Family: PF_INET Service: SOCK_STREAM Local IP: Local Port: Remote IP: Remote Port: Family: PF_INET Service: SOCK_STREAM Local IP: Local Port: Remote IP: Remote Port: 44287

 The domain parameter of socket() specifies the protocol family  PF_INET : IPv4 Internet protocols  PF_INET6 : IPv6 Internet protocols  PF_UNIX / PF_LOCAL : Local communication  etc.

 The type parameter of socket() specifies the communication semantics  SOCK_STREAM : Connection-oriented, sequenced, reliable, two-way communication of byte streams  SOCK_DGRAM : Connectionless, unreliable communication of datagrams (messages of fixed length)

 Use get/setsockopt() to manage options on an existing socket  SO_ACCEPTCONN  SO_BROADCAST  SO_DONTROUTE  SO_ERROR  SO_KEEPALIVE  SO_LINGER man 7 socket

 HTTP is the protocol for communication between browser apps and Web servers  Web servers are essentially HTTP servers  Protocols have versions  Most clients and servers support version 1.1  But 1.0 is also in use (maybe also 0.9?!) why?

 Each layer prepends or appends its information in a header or trailer P Ethernet Hdr | IP Hdr | TCP Hdr | HTTP Request | Cksum IP Hdr | TCP Hdr | HTTP Request TCP Hdr | HTTP Request HTTP Request

PQ

 RFC 1945 is the HTTP 1.0 standard  see  RFC 2616 is the HTTP 1.1 standard  see  RFC 2396 is the URI standard  see

 From the RFC:  HTTP is an application-level protocol with the lightness and speed necessary for distributed, hypermedia information systems

 Again from the RFC:  HTTP communication generally takes place over TCP/IP connections  The default port is TCP 80, but other ports can be used  HTTP is not dependent on a specific transport layer https is typically TCP port 443

 HTTP defines a very simple structure:  A client sends a request  The server sends a response  HTTP supports multiple request/response exchanges over a single connection  e.g. try using telnet to access a Web server....

 HTTP requests are line-based ASCII text  Lines must always end with "\r\n" (a.k.a. CRLF )  Headers are optional  A blank line separates the request from the content Request-Line Header(s)... Header(s) blank line -- Content... Content... what content?!

 The Request-Line consists of 3 tokens:  Each token is separated by a space character  Though "\r\n" is required by the protocol, "\n" seems to work in practice  The HTTP-Version is either HTTP/1.0 or HTTP/1.1 Method URI HTTP-Version\r\n

 The HTTP request’s Method can be:  GET – request information identified by the given URI (absolute or relative?)  HEAD – request metadata regarding the given URI (search engines!)  POST – send (i.e. post) information to the given URI (e.g. via a form) Method URI HTTP-Version\r\n

 The HTTP request’s Method can be:  PUT – store information in the location identified by the given URI  DELETE – remove the entity identified by the given URI (really?) Method URI HTTP-Version\r\n

 The HTTP request’s Method can be:  TRACE – used to trace HTTP forwarding through proxies, tunnels, etc.  OPTIONS – determines the capabilities of the Web server or the characteristics of the named resource Method URI HTTP-Version\r\n

 The GET, HEAD, and POST methods are supported everywhere  Check out homework #2!  HTTP 1.1 servers might support the PUT, DELETE, TRACE, and OPTIONS methods (but not always!) Method URI HTTP-Version\r\n

 The URI is defined in RFC 2396RFC 2396  An absolute URI consists of four parts:  A relative URI omits the scheme and server: ▪ The server is assumed (since we’re already connected) scheme://hostname[:port]/path /path which one should we use in our HTTP Request-Line ?

 In general, relative URIs are used in the HTTP Request-Line  HTTP 1.1 servers are required to support absolute URIs, but not all do  When using a proxy HTTP server, an absolute URI is required  Or else, the proxy server won’t know where to find the resource (i.e. document)

 After the Request-Line, the request might have header lines  Header lines specify attribute name/value pairs (e.g. User-Agent: )  Note that HTTP 1.1 requires the Host: header always be included Request-Line Header(s)... Header(s) blank line -- Content... Content...

 Request headers provide information to the server about the client  Who is making the request  What kind of client is making the request  What kind of content will be accepted  In HTTP 1.0, all headers are optional  In HTTP 1.1, the Host: header must be sent

 Headers can be included in any order:  For GET and HEAD requests, that’s the end (though don’t forget the blank line!) GET /index.html HTTP/1.1 Accept: text/html Host: From: User-Agent: Mozilla/4.0 Referer: Accept: text/html Host: From: User-Agent: Mozilla/4.0 Referer: -- blank line --

 If a POST request is made, the headers must include Content-Length : POST /~goldsd/changegrade.php HTTP/1.1 Accept: */* Host: User-Agent: SecretAgent v3.0 Referer: Content-Length: 36 Accept: */* Host: User-Agent: SecretAgent v3.0 Referer: Content-Length: blank line -- rin= &item=midterm&grade=104

 HTTP responses are line-based ASCII text  A Status-Line is always returned  A blank line separates the response from the content  Content is a sequence of bytes (e.g. HTML, image, text, etc.) Status-Line Header(s)... Header(s) blank line -- Content... Content...

 The Status-Line consists of 3 tokens:  The HTTP-Version is either HTTP/1.0 or HTTP/1.1 (and does not necessarily match the corresponding request)  Response status is represented using a 3-digit Status-Code and a human-readable Message HTTP-Version Status-Code Message

 Status codes are grouped as follows:  1xx – Informational  2xx – Success  3xx – Redirection  4xx – Client Error  5xx – Server Error (click me)

 Example status lines include:  HTTP/ OK  HTTP/ Moved Permanently  HTTP/ Bad Request  HTTP/ Forbidden  HTTP/ Internal Server Error

 After the Status-Line, the response typically has header lines  Header lines specify attribute name/value pairs (e.g. Date: )  As with request headers, response headers end with a blank line Status-Line Header(s)... Header(s) blank line -- Content... Content...

 Response headers provide information to the client about the entity (i.e. document)  What kind of entity/document  How many bytes are in the document  How the document is encoded  When the document was last modified  The Content-Type header is required, as is the Content-Length header (usually)

 Headers can be included in any order: HTTP/ OK Date: Wed, 30 Jan :48:17 EST Server: Apache/1.17 Content-Type: text/html Content-Length: 1756 Content-Encoding: gzip Date: Wed, 30 Jan :48:17 EST Server: Apache/1.17 Content-Type: text/html Content-Length: 1756 Content-Encoding: gzip -- blank line fjfjef0jefe0fje2f0je2f0je2f0e2jfe0fje20 fj2e0fjef0jef0e2jf0efje0fje02fje20fje2f0ejf 0jef2e09fj209g209fj20gag09ha0gh0agha0gjg0jg

 For HTTP 1.0, default behavior is as follows:  Client sends a complete HTTP request  Server sends back a complete HTTP response  Server closes its socket  Therefore:  If the client needs another document (e.g. images, CSS, etc.), the client must open a new socket connection!

 In HTTP 1.0, support for persistent connections is available  Multiple requests can be handled over a single TCP/IP socket connection  The Keep-Alive: header is used to keep the connection alive

 As of HTTP 1.1, support for persistent connections is available (and is the default)  Multiple requests can be handled over a single TCP/IP socket connection  The Connection: header is used to exchange information about persistence ▪ e.g. Connection: close