Charlie Garrod
--------------

A good systems paper needs to clearly answer the following questions:

1.  What did the authors do that is worth reporting in a paper?  Is their
work primarily theoretical?  Did they design a new system, implement a
system, or analyze an existing system?  If it isn't clear what the authors
actually did, it's probably not worth reading a paper about it.

2.  What new ideas are being presented?  How do these ideas relate to 
other proposed solutions to the problem?  What advantages and disadvantages
does the authors' approach have compared to previous systems?  To discuss
this well, the authors need to have extensive knowledge of previous related
work in their area, as well as a deep understanding of the fundamental 
problems addressed by their work. 

3.  What choices did you encounter while doing your work?  What decisions
did you make, and what motivated you to make those decisions?  What
consequences would there be had you chosen the other alternatives?  
Basically, if another group were to mimic your work by building a similar
system, your paper should function as a roadmap for them, outlining what
choices they will have to make and including your analysis of those choices.

4.  Finally, what did you learn from your work, and more importantly, 
what should the reader learn from your paper?  Fundamentally, what 
contributions have you made to the field?  If the reader could only take 
home one piece of critical information from your paper, what would it be? 
What implications does your work have for the design of similar systems?
What assumptions does your work make?  


To address each of these issues, a good paper should have the following 
sections (not necessarily in this order):

1.  An introduction, to clearly state what problem the authors are
attacking, why it is important, and quickly state what they did.

2.  A section discussing related work, putting the authors' work in 
context of what has been done before.  

3.  A section describing what the authors did in detail, including
the rationale for decisions that the authors made.

4.  An analysis of the authors' work, including an analysis of potential
competing designs. 

5.  A conclusion, quickly summarizing the results of the authors' work
and reiterating any important messages to be learned from the paper.


Most of the papers we have read this semester have one shared strength:  they
each present an idea or ideas that are sufficiently new, to the extent that
the idea itself makes a significant contribution to the field.  To be great,
however, the paper must present the ideas in a clear and concise manner.
For instance, I feel that the XFS paper did not succeed in this goal.  Rather
than describe the fundamental contributions that XFS made to the field, the
paper spent most of its time descibing the minute details of the system,
effectively drowning out the more important information with insignificant
details.  While it's great that I understand how XFS works, its unfortunate
that I can't clearly describe how XFS is important, and therefore to me at
least, XFS is not important.