Dynamic linking is ubiquitously used, but it's hard to find
a detailed, accurate description of how it works. Nevertheless,
it is a remarkable example of advanced software engineering,
and no modern system is possible without it.

Why Solaris abandoned static linking:
http://blogs.oracle.com/rie/entry/static_linking_where_did_it

== Reading

Read the "Dynamic linking and loading" chapter in John Levine's
"Linkers and Loaders" (http://www.iecc.com/linker/). Concentrate
on ELF linking and loading, but have a look at the Microsoft  
section if you are curious.

Read your dynamic linker's man page. On Linux and Solaris: ld.so

== Addtional reading:

"Cheating the ELF. Subversive Dynamic Linking to Libraries" by the grugq. 
http://althing.cs.dartmouth.edu/local/subversiveld.pdf

"Modern Day ELF Runtime infection via GOT poisoning" by Ryan O'Neill
http://vxheavens.com/lib/vrn00.html

== Suggested homework:

Use a dynamically compiled executable on your system (which, say,
prints "Hello world" in a loop 10 times) and a debugger to observe
the stack and GOT just before and after dynamic linking. 

What are the arguments that the dynamic linker entry point receives
when entered through GOT[]? What else is stored in GOT?

Observe how the symbol to be linked is referenced from the PLT stub.

Useful GDB tips:
http://www.unknownroad.com/rtfm/gdbtut/gdbadvanced.html

Remember that /proc/<PID>/maps will show you the process' memory
layout with all of its currently mapped code, libraries and stack.