I often hear the advice, "Read the masters" (i.e., read old, classic texts by great mathematicians). But frankly, I have hardly ever followed it. What I am wondering is, is this a principle that people give lip service to because it sounds good, but which is honored in the breach more than in the observance? If not, which masterworks have you found to be most enlightening?
To keep the question focused, let me lay down some ground rules.
List only papers/books from the 19th century or earlier. I recognize that this is an arbitrary cutoff but I want to draw a line somewhere.
It must be something that you personally have read in its entirety (or almost in its entirety). I'm not really interested in secondhand evidence ("So-and-so says that X is a must-read").
You must have acquired important mathematical insights (not just historical insights) from the paper/book that you feel that you would never have acquired had you restricted your reading to 20th-century and 21st-century literature. It's not enough, for my purposes, that you found the paper/book "interesting" but not really essential. If possible, briefly describe these insights in your response.
[Edit: In response to a comment that suggested that I have set the bar impossibly high, let me violate one of my own ground rules and point to this discussion on the $n$-Category Cafe that gives some secondhand examples. That discussion should also help to clarify what I am asking for more examples of.]