Writing Advice






What Will You Be Writing and When?

·	 Write early and often

·	 Helps to focus logic

·	 Great mechanism to share ideas

·	 Technical reports:  good ideas, preliminary results, 
early versions of papers

·	 Conference papers:  varying difficulty of acceptance;  
tough conferences typically have 20% to 25% 
acceptance rates

·	 Workshop papers:  often easier to get accepted, great 
opportunities to meet important researchers in 
focused areas

·	 Journal papers:  typically more content than 
conference papers, approx. 2 year turn-around

Barriers to Writing

·	 Writing well is hard

·	 Fear 

·	 Lack of consistent effort

·	 Poor writing skills

·	 Little Red Schoolhouse course

·	 Every quarter

·	 Great for non-English speakers

·	 Good review of basic grammar, sentence 
structure, usage

Do You Need a Review of Basic Writing?

·	 Are there errors in the following sentences?



·	 We used a defective hammer which broke after 
pounding the first nail.

·	 The dog lost it's bone.

·	 He had to choose between three options.

·	 I feel badly about that.

·	 She went along with Emily and I when we visited 
Mom.

·	 After our company hired a new typist, our proofreader 
found less mistakes than usual.

Typical Paper Organization  (Systems)

·	 Abstract:  summarize results

·	 Introduction:  motivate problem, again 
summarize results,  outline of paper

·	 Background:  discuss issues, design choices

·	 Experimental methodology:  describe 
simulation environment, testing conditions, 
benchmarks, hardware, experiments

·	 Goal:  describe experimental conditions in 
enough detail that other researchers can 
duplicate your experiments

·	 Results:  experimental data, analysis

·	 Conclusions:  summarize results, contributions 
of paper

Strategies for Writing

·	 Start with an outline

·	 Write rough drafts



·	 Be concise (hard)

·	 Two extremes:  too much or too little



·	 State ideas clearly (harder)



·	 Avoid the wrong tone:  not too formal, not too 
colloquial  

·	 (e.g., don't use Usenix-style slang about hacks in 
a conference paper;  know your audience)

·	 Finding the right tone:  
read lots of papers!  (good ones)

·	 Avoid unnecessary jargon 



·	 Define terms the first time they are used

·	 Sometimes need to re-define later in the paper



·	 Write in third person plural ("We" not "I")



·	 Use active voice, not passive voice

·	 "We designed and measured a large disk array" 
NOT "A large disk array was designed and 
measured"



·	 Start paragraphs with topic sentences



·	 Overall structure:  emphasize main ideas and results



·	 Chronological order is not the best 
(First I did X, then I did Y, ....)



·	 Just because you spent a lot of time working on 
something doesn't mean it contains an 
interesting idea  



·	 "Avoid adverbs"

·	 Eliminate "very", "seriously",  "extremely"

·	 Don't say your technique is "very effective"-- 
state your results and let others evaluate them

·	 Technology trends change fast:  "very high 
bandwidth" is likely to look low in a few years



·	 Write smooth transitions between sections

·	 Don't be arrogant

·	 "We invented this technique..."

·	 "We discovered this fundamental truth..."



·	 Don't write a mystery novel

·	 Don't put minor points before major ones (This 
method was good, .... but a second method was 
better, .... and the third was best of all.)

·	 State main ideas and results in abstract and 
introduction

·	 Preview results at beginning of each section

·	 Clearly summarize results in conclusion section

·	 "Tell them what you're going to tell them, tell 
them, and tell them what you told them."



·	 Think about what a casual reader will look at:

·	 Abstract

·	 Introduction

·	 Figures/tables

·	 Write captions that explain the lessons to be 
learned from the graphics

·	 Conclusion



·	 Editing tips:

·	 Read it out loud

·	 Spell check

·	 Grammar check

·	 That/which checks

·	 Eliminate adverbs ("very")

Questions from "How to Write a Good 
Systems Paper"

·	 Original Ideas

·	 Are the ideas in the paper new?



·	 How do you know?  Know the state of the art, current 
research



·	 Can you state the new idea concisely?



·	 What is the problem being solved?  Explain why it 
can't be solved using existing techniques



·	 Is work significantly different from other published 
work, or an obvious extension of  existing 
techniques?  (avoid LPU:  least publishable unit)

·	 Are the ideas significant enough to justify a paper?



·	 Are comparisons with previous work clear, explicit?



·	 Does the work comprise an extension, validation or 
repudiation  of earlier but unproven ideas?



·	 Have you referenced older as well as recent papers?



·	 Don't "reinvent" 20-year-old ideas

·	 Avoid unpublished material, personal 
communication references



·	 Goal:  other people can follow your reasoning, 
have access to the same background material 
you had

·	 Reality

·	 Be honest:  implementation or paper design?



·	 If implemented, how has it been used and what has 
been learned?



·	 Negative results are interesting:  "we thought this 
would work, but it didn't"



·	 Usual model:  publish a design paper early, and then a 
year or two later, publish an experience paper



·	 Usually:  successful design papers include early 
implementation results







·	 Lessons Learned

·	 What have you learned?  

·	 What should readers learn?  Spell lessons out.

·	 How generally applicable are these lessons?  State 
assumptions on which conclusions rest.



·	 Choices

·	 What were the alternatives considered?

·	 Why were choices made the way they were?  Explain.

·	 Were original choices correct?  Were hypotheses 
correct?







·	 Context

·	 What are the assumptions?

·	 Are they realistic?  Especially important for 
unimplemented systems.  (Simulators only as good 
as the models on which they are based.)

·	 How sensitive are results to the assumptions?



·	 Focus

·	 Remove excess baggage from introductory sections

·	 Support your main points

·	 Include just enough material from earlier work for 
readers to follow arguments





·	 Presentation

·	 Are ideas organized and presented clearly, logically?

·	 Terms defined before use?

·	 Minimize forward references  ("Each file contains a 
list of items, which will be described in Section 9")

·	 What is intent of paper:  global survey or selective 
treatment

·	 Does abstract communicate important ideas of paper?

·	 Avoid passive voice

·	 Avoid references to future work.  ("Note to reviewers:  
this vaporware design will be fully implemented, 
debugged, measured and evaluated in the final 
paper."  ---  Reviewers don't believe these!)

Thesis Writing

·	 Consistent effort

·	 Avoid burnout

·	 Pick your most productive hours of the day

·	 Guard those hours vigilantly:  Write every day 
during that period

·	 Your job:  2 to 3 hours of writing every day