UMBC CMSC 313, Computer Organization & Assembly Language, Spring 2002, Section 0101

Project: Base64 Content-Transfer-Encoding

Due: Tuesday March 12, 2002

Objective

Background

Exchanging binary files by email is not quite straightforward because many mail servers were designed to handle text, not binary data. Attempts to send binary files through these servers can result in mangled files. For example, some mail servers might ignore the most significant bit of each byte, since standard ASCII encoding uses only 7 bits. Other mail servers truncate all data beyond the 80th character of each line. In fact, the whole concept of a line is meaningless when we work with binary files. To complicate matters, email is often routed through several servers, so the problem might not be with either the sender's mail server or the receiver's mail server.

The MIME (Multipurpose Internet Mail Exchange) standard defined in Internet RFC 1521 is a comprehensive mechanism for formatting Internet messages. For many people, MIME is synonymous to email attachments. We are interested in just one section of this standard the Base64 Content-Transfer-Encoding that specifies how binary files should be converted into a text file that can be sent intact through most mail servers. The complete specifications of the Base64 standard are (what else) attached at the end of this project description.

Remark: Unlike other organizations (e.g., ANSI, ISO) which publish standards with lofty-sounding titles, the Internet Engineering Task Force's (IETF's) standards are for historical reasons published as Request for Comments (RFCs). Although not all RFCs are standards, the specifications of just about every Internet protocol can be found in an RFC. For more information on RFCs and how they are published, check out http://www.rfc-editor.org.

Assignment

As a reference standard, we will use the mimencode command on linux.gl.umbc.edu. Using mimencode with the -u option, we can convert the output of your program back to binary. If your program works correctly, the output of mimencode -u should be identical to the original input file. For 15% extra credit, write an assembly language program that reverses the process of your first program. I.e., the second program prompts the user for an input file and an output file. If the input file is a properly formatted text file that conforms to the Base64 standard, your program should store the corresponding binary file in the output file.

Implementation Issues

1. All of the file conversion must be done by your program. You are, of course, not allowed to make a system call to mimencode.
2. Files can be opened for reading using a system call to the open() function. The C function prototype of open() is:

         int open(const char *pathname, int flags);
   

   According to the Linux system call convention, the syscall number for open() should be stored in EAX, a pointer to a null-terminated string with the name of the file to be opened should be stored in EBX and the flag O_RDONLY should be stored in ECX. The return value, stored in EAX, is a file descriptor (a 4-byte integer) that can be used in subsequent syscalls to read(). Further information on open() can be obtained from the Linux man pages. Type 'man 2 open'.
3. Symbolic constants for syscall numbers, flags, etc can be found in a file called stddefs.mac in the directory: afs/umbc.edu/users/c/h/chang/pub/cs313. Copy this file into your own directory. Then, the file can be included in your assembly language program using the NASM directive:

         %include "stddefs.mac"
   

4. To open a file for writing, a syscall to creat() is more appropriate. The C function prototype for creat() is:

         int creat(const char *pathname, mode_t mode);
   

   Calling creat() is very similar to calling open(). The difference is that the file is opened for writing and the file is created if it does not already exist. If a file with the same name already exists, it is overwritten. As before, the return value stored in EAX is a file descriptor. The second argument to creat() is used to set the permissions of the newly created file (as in the chmod Unix command). You will most likely want to allow the user to read and write to the file, so store the expression S_IREAD|S_IWRITE in the ECX register. S_IREAD and S_IWRITE are defined in stddefs.mac.
5. Remember to close all open files before your program terminates. This is accomplished with a syscall to close() with the file descriptor as the sole argument. The close() function has the following function prototype:

         int close(int fd);
   

6. Once a file is opened, you can read from and write to it using the read() and write() syscalls as you have done with stdin and stdout.
7. Despite what the man pages say, you can tell that you have reached the end of a file you are reading when read() returns 0.
8. Recall that read() stores the characters read at the address provided and returns the number of characters read. The string read in is not null-terminated. Also, if the string is read from stdin, the last character is a '\n'. Thus, some massaging of the string is needed before it can be used as a file name.
9. You should not assume that the file has run out of bytes when read() does not return the maximum number of bytes requested.
10. It is inefficient to read 3 bytes at a time.
11. The functions open(), creat() and read() return the value -1 if an error is encountered. The cause of the error is given as an error code in the global variable errno. If you wish to examine these values, you must declare errno to be an external label. Symbolic names for some of the possible values for errno can be found in stddefs.mac. Consult the Linux man pages for the meaning of each error. If you reference errno, then you must link your program using 'gcc -nostartfiles' instead of ld.
12. Recall that the Intel Pentium CPU is little endian. If you move multiple bytes into a register, the bytes might not be ordered the way you like.
13. Assembly language instructions that you might find useful include: AND, OR, SHL, SHR, XCHG.
14. A common task that you will want to perform is: add a new character to the output buffer, then write out the buffer if it is full. You will probably want to write a subroutine to do this. Invent your own parameter passing conventions.
15. Read the Base64 specifications for handling the last few bytes of input carefully. The output may need to be padded with 1 or 2 '=' as appropriate.
16. If you want to have your output appear identical to the output from mimencode, print out 72 characters per line.

Turning in your program

Use the UNIX 'submit' command on the GL system to turn in your project. The class name for submit is 'cs313' and the project name is 'project'. Sample runs and a typescript file is not needed for this project. The grader will simply test your program using mimencode and some binary files. Include a README file if your submission needs any special attention.

References

1. Borenstein, N. and Freed, N. “MIME (Multipurpose Internet Mail Extensions) Part One: Mechanisms for Specifying and Describing the Format of Internet Message Bodies.” RFC 1521, September 1993. Available at ftp://ftp.isi.edu/in-notes/