Some more comments on your report 
"Implementation of Interactive New York State Terrain Viewer"
for your independent study

In Table 1, I assume that the second row should be 1000x1000, rather than
just 1000. As well as measured download times, I would like to see some
theoretical download times, at a rate of 1 Mb/sec (which you might expect
to get over Ethernet), 128 kb/sec (ISDN), and 28.8 kb/sec (modem). 

For the results in Table 1, I am surprised that the image creating time is
so large - 12 seconds for a 600x600 pixel image. What exactly is involved
in "image creating" and "image drawing"? I thought that the Java 
implementation in Netscape used native methods to do the JPEG decompression
etc, so downloading a JPEG image and displaying it using Java shouldn't be
much slower than downloading and displaying it in Netscape. I would be 
interested in seeing some comparisons in times to do this for different
sized images.

In Table 2 and Table 3, you compare sizes and times for one large image
(600x600 pixels) and 4 image chunks (300x300 pixels each). The compression
times are the same for both, which I don't understand, since I would 
expect that since the chunks are 4 times smaller, they would be compressed
about 4 times faster.  

Also, I would like to see some timings and data sizes (like those in the 
tables) for the actual chunk sizes that you use, i.e. 160x160.

One thing you might want to check to try to speed up the program, is to try
to reduce as much as possible the copying of data, i.e. try to move the
objects rather than copy them whenever you can. When Animesh was developing
the Tcl/Tk version of the 2D terrain viewer, he said that he greatly improved 
the speed by moving pointers to the chunks rather than copying chunks from 
buffers into image arrays.