The question is simple: what is the relationship of patterns and architectures?

The hypothesis is also simple: you can derive any architecture by the successive application of patterns. The paper tries to convince the reader that this is true by deriving a fairly complex architecture, HotDraw, solely by applying patterns. 

I expected the paper to spawn at least one PhD proving or disproving the thesis, and a literature of PgaDescriptions of architectures. Ralph and I still talk about a "PGA description" of a system, but nobody else does. --KentBeck

----
The story:
Early in the public life of SoftwarePatterns people were asking, "You talk about patterns, and frameworks, and architectures, and they all seem really similar and really different. What is going on?"

One evening at the ThornwoodWorkshop, KentBeck and
RalphJohnson hid away in a corner to derive HotDraw
from patterns.  Kent was sure it could be done.
Ralph hoped it could be done.  Kent was right, and
after a few hours they had both an outline of the
derivation as well as some new patterns.

Kent typed the outline into his Mac and described some of the patterns.
Over the next few months the paper went back and forth
over the Internet, until it was submitted for ECOOP'94.
It was accepted, and you can get the PostScript, PDF, ... for it: http://citeseer.nj.nec.com/27318.html