Last updated on March 24

Exercise 2

In this exercise you will write a simple web server. The only purpose of this server is to provide http interface to DNS service.

Content:

Web Server
Your Program
Bonus
The zone's primary server
Library Usage Limitations
Non-blocking mode
Testing Your Programs
Error Handling
Relevant RFCs
Relevant Man Pages:

General

Web Server:

The http DNS gateway will hereafter be called 'httpDNSgateway' and can get either the first one or all the three of the following parameters:

A port for receiving http requests
A machine running the primary name server (optional)
A primary name server's port (optional)

For example, httpDNSgateway can be invoked like this:

httpDNSgateway 2434 shuldig.cs.huji.ac.il 53 or
httpDNSgateway 2434

Your Program

If httpDNSgateway was started with only one parameter it takes the primary dns server's machine from /etc/resolv.conf and uses port number 53.
httpDNSgateway runs forever (it is supposed to be a daemon).
httpDNSgateway should accept http queries on the specified port and send the http reply back to the requester.

'httpDNSgateway' accepts HTTP requests from clients on port < portnum >. Your 'httpDNSgateway' supports only requests generated by web browsers upon form submission. (A form sample can be found here). Note that only the method GET is to be supported. In general, requests generated by web browsers upon submitting GET based forms have the following structure:

	GET <B> /<resolver>?<DATA> <B> HTTP/1.0<CRLF>
	Referer: ... <CRLF>
	Connection: ... <CRLF>
	User-Agent: ... <CRLF>
	Accept: ... <CRLF>
	<CRLF>

You have to deal only the first line. You may ignore all the others.

The tokens have the following meaings:

<resolver> is the name of the object which tells your http server to translate the host name to ip.
<DATA> is a string in the form
<field1-name>=<field1-value>&<field2-name>=<field2-value>& ...,
where
- <field-name> is a form field name (e.g., "user").
- <field-value> is a value typed by a user in the corresponding form field.
- Each space in <field-name> or <field-value> is substituted by '+'.
<data-length> is a length of data in a POST based request.
<B> is a space
<CRLF> stands for "Carriage Return and Line Feed", i.e., it is equivalent to the "\r\n" string in C.

Whenever httpDNSgateway receives a request, it contacts the primary DNS using UDP/IP, and finds the information there.

httpDNSgateway sets the "recursion desired" bit in the DNS request in accordance with the information provided by user in the html form.

httpDNSgateway is supposed to support queries only. It must support only type A queries (host addresses), of class 1 (Internet). Inverse queries and wildcards are not to be supported.

httpDNSgateway is supposed to support only one request per connection and only one question per client message.

httpDNSgateway uses UDP/IP to communicate with the primary server and must be able to handle communication failures (message losses): httpDNSgateway retransmits the requests to the primary server up to 5 times (timeout between retransmissions is 5 seconds).

httpDNSgateway must be client-misbehavior proof. This means that httpDNSgateway must not crash if a client misbehaves (e.g. sends wrong messages, closes connection in the middle of sending the message, does not read from the socket). httpDNSgateway should close the connection of misbehaving clients. If a client is idle for more than 2 minutes, the corresponding socket is to be closed.

Bonus #1: 5 Points

httpDNSgateway must support multiple connections concurrently (i.e. a second client should not be blocked from connecting to the server while the first client's request is pending). When multiple clients request the same name simultaneously, httpDNSgateway does not send multiple queries to the primary server, but rather waits for the reply from the server and then sends the answer to all the requesting clients.

The zone's primary server

The domain name server running on the cs network is (see /etc/resolv.conf): shuldig.cs.huji.ac.il

The standard port (53) is used for servicing DNS requests, using both UDP/IP and TCP/IP (we will use UDP/IP). For example, when you use gethostent call within a program, the corresponding request is sent to the name server.

The primary zone has all the local information for mapping local machine-names to IP addresses. In addition, the primary server knows how to contact higher-zones servers in case it does not have the requested information.

Library Usage Limitations

Your httpDNSgateway is explicitly forbidden to use any existing library functions/tools/utilities that handle DNS queries.
However, you may use these utilities for debugging/testing purposes.

Non-blocking mode

httpDNSgateway must be client-misbehavior proof, therefore, it must operate in non-blocking mode. httpDNSgateway must take into consideration that system calls may be interrupted by signals and have to be repeated.

Testing Your Programs:

You may test your programs by submitting the form. A sample form can be found here. You have to change ip and port number in form.html in order to have the browser send the request to your program. The form can be submitted using your favorite web browser (e.g., Netscape).

You can also connect and "talk" to httpDNSgateway using telnet. Another option is to use wget utility.

You can compare obtained query results with those of UNIX utils dig, host or nslookup utilities.

Error Handling:

In all cases of errors, your program must print an informative error message to stderr (use perror when needed).
In case of an unrecoverable error (e.g., if 'httpDNSgateway' can not open the server socket), the program must exit with exit status 1.
If a client's request cannot be fulfilled, 'httpDNSgateway' sends http response to the client stating the reason. For example 404 error code indicates that the host can not be found. 504 error code is used to indicate that the primary dns server does not respond.

Relevant RFCs:

RFC 1945.Hypertext Transfer Protocol -- HTTP/1.0
RFC 2616.Hypertext Transfer Protocol -- HTTP/1.1
RFC 1866. Hypertext Markup Language - 2.0.
RFC 1034. Domain names - concepts and facilities
RFC 1035. Domain names - implementation and specification

Relevant Man Pages:

socket(2), bind(2), listen(2), accept(2), close(2), shutdown(2), connect(2), send(2), recv(2), read(2), write(2), fcntl(2), htons(3), gethostbyname(3x), tcp(4), udp(4), sendto(2), recv(2), select(2), perror
You may use HEADER structure from /usr/include/arpa/nameser_compat.h

General:

Submit your sources, a README and a Makefile that prepares httpDNSgateway.
Follow the general instructions in ~com1/WWW/guidelines.html.