CSR Tales
Published in

CSR Tales

CSR Tale #3: The paper that took 10 years to publish!

CSR Tales goes international in this post, with our third tale coming from Prof. Javier Bustos-Jiménez in Chile. Javier is the director of the NIC Labs at the University of Chile. His research interests include complex networks, internet protocols, network privacy/security, and data science. His “not-research interests” (his words) include sports, music, and coffee. His Erdős:Bacon:Sabbath number is 4:4:4. Javier tells us the thrilling story of a particular paper that took ten years to get published!

In Javier’s words:

My story revolves around the the development of a new disruptive technology on camera chips that works in conjunction with a different operational (CPU) chip by SMaL technologies. At the beginning, the SMaL video camera had a typical CPU for embedded technologies with only 500 Kbytes of RAM. The camera sent 640x480 color JPEG frames (having usual sizes of around 50 Kbytes each) to a maximum of 20 clients.

How do I know it?

Because I worked on how to increase the number of clients without a
quality drop.

In 2002 I was finishing my master thesis at the computer science
department (DCC) of Universidad de Chile
, working in new text indexes
for non-exact queries with Prof. Gonzalo Navarro. I was working
in Information Retrieval even though my research/engineering skills
were more related to System Engineering mostly because the former
was (still is) the strongest area of the department. Moreover, being
one of the few students working in systems gave us some benefits, like access to of the office of a senior professor — José (Jo) Piquer (I was his TA in networks and systems courses) — while he was directing the department.

Suddenly, Prof. Piquer came to his office and told me:

Jo: We have an interesting problem here, the team behind SMaL cameras have come to us because they need to increase the number of served clients
on their internet cameras.

Me: ?

Jo: They are selling a surveillance camera connected to Internet, the main problem is they are losing quality with around 20 clients.

Me: It looks like the typical problem of readers-writers, does it not?

Jo: Yes, but in this case I think we have to find a way to group clients by their download speed.

We then developed a camera frame buffer array where each buffer stored a JPEG frame with its timestamp and the number of clients reading it, and tested frame replacement algorithms for the camera and a frame selection algorithms from the client threads (just like a shared-memory buffer). As a result, with 4 buffers we served 6 clients without degradation, 18 clients with 6 buffers, 50 with 8 buffers, etc. SMaL Technologies were so happy with our findings that they gave us funding to continue research in whatever we wanted.

I told Jo that we have to publish our work, that it was a clean and beautiful solution for a real problem. So, our first thought was to send it to a mid-league conference: the national (then regional) conference on computer science SCCC 2003 (in those days, a very important venue for all Latin-American researchers, with its deadline around April), I still remember the comment from the reviewer that gave us a reject:

I cannot see any academic contribution on this work

We first were shocked, then worried, “How could we be so wrong?”. Maybe the SCCC was not the venue for our work; in that time it started to focus on two main areas: software engineering and programing languages, and systems was left behind given the low number of academic researchers in the area.

In September 2003 I started my PhD at INRIA Sophia-Antipolis (France), so we waited until 2004 for a huge revision of the article. We discussed that maybe the problem + the algorithms + the demo were not enough for an academic venue. I used SciLab (an open-source version of MatLab) to model the worst-case scenario: where all clients ask for frames periodically with a delay of prime numbers milliseconds (so they will never coordinate). The simulation was consistent with our demo so we were happy and confident that the work will be accepted anywhere. We submitted to the IEEE Transactions on Multimedia Journal. We were sure it would be accepted.

The decision: REJECT

But… why? We received the reviews in November 15, 2005. First reviewer loved our work, second reviewer… well: “The algorithm proposed in this paper is a standard practice used in Operating Systems”. We knew that! That’s why we applied it on this problem! But the last stone in our grave was given for the third reviewer, in a practice that I have seen in all my years in academy, and it is called “the snobbish non-expert reviewer”, who started with “Although I am not really an expert on the contents of the paper…” (so I reject it). We still don’t know they did not refuse the review if they were not an expert.

Anyway, first reviewer pointed that maybe there will be some issues in the synchronization of the delivered frames. We noticed that we did not explain that part clearly so we did some minor corrections on the next draft, and added a new section about the perceived playback speed on the client side (we worked on QoE before it was cool ;-) ), an idea I discussed with a bunch of PhD students when I attended SIGMETRICS 2016 in Saint-Malo, France (the truth is the discussion was performed while I watched the 2006 World Cup in a bar, sharing beers and frites). Now the article was ready and it should be accepted anywhere. Also, I received my diploma at the end of that year!

The paper was submitted to a regular issue Elsevier’s Journal of Visual Communication and Image Representation in 2007, and it was rejected because the work was not visual communication nor image representation (our mistake!).

Then, it was submitted to EuroSys Conference 2008 as an example of a
working demo of a real system, and it was rejected because the work
was “too theoretical”.

We revised the article and submitted it to ACM Multimedia Conference
(2009), and it was rejected because it was “too practical”.

With more experience publishing articles in conferences, I told Jo Piquer: “We are looking in the wrong area, this is not a multimedia work, it is a classical systems solution, so we have to send it to a conference in that area”. The selected conference was Euro-Par 2010 [*] (on those days a class A conference), the article was accepted and we were running for the best paper award until somebody presented a whole study about people panicking in public places (like cathedrals and malls) using game engines for the simulations. You can find the paper here
https://link.springer.com/chapter/10.1007/978-3-642-15277-1_53

[*] I’ve selected that conference because we (Jo and I) were part of an European Network of Excellence that organized lot of meetings between 2004 and 2008, and the participants of those meetings were also in the Euro-Par community. First thing I’ve learnt in academy: lobbying is as important as a good work.

My 5 cents from this experience are:

  1. Never give up a good work.

2. Reviewers are humans, and as humans they are prone to mistakes,
even more if they don’t know you.

3. A good venue selection increases the chances of a work.

4. Never miss the chance to discuss your work with some fellows, specially if they come from another background, they could give you some insights you are missing.

5. About point 4, now it’s easier with the use of social networks, you don’t have to wait for conferences… or world cups.

Editor’s Notes. I want to highlight a few things in Javier’s tale.

  1. Bad reviewers are everywhere. Read Steve Swanson’s excellent blog post about how to write papers such that they get past these reviewers. I’ve personally seen reviewers who have poor expertise, yet will have strong opinions and go on at length at the program committee meeting. It is the job of the Program Committee Chairs to weed them out, or once they are in the committee, to minimize the problems they cause. Program Committee Chairs who try to be nice to everyone are doing the community and the program committee a disservice.
  2. Where you submit your work is crucial. What junior grad students do not know is that each conference has its own community with different values. For example, SOSP/OSDI has a strong focus on evaluation: weak evaluation will get your paper rejected, even if it would been accepted at some other conference. It is crucial to learn these nuances quickly; your advisor is a good person to talk to about this!
  3. Discussion makes research stronger! Never miss a chance to talk about your research to other professors, your fellow grad students, and visitors. Personally, submissions on which I’ve gotten feedback prior to submission, have had a higher acceptance ratio than submissions on which I haven’t. The more feedback a submission gets, the better the chances of acceptance!

--

--

--

Documenting the stories behind computer science research. We bring you the tales behind how different research projects got started. We highlight the social aspect of doing great research, and how luck is often involved.

Recommended from Medium

5 Reasons to use GraphQL and Prisma

Understanding Dapps

Undefined symbols and where to find them

Koinos: Now with Google’s Protobuf!

Lenovo ThinkPad X1 Carbon 14" G6 6th Gen Intel Core i7–8650U 16GB RAM 1TB SSD

Lenovo ThinkPad X1 Carbon 14

Setting up mTLS and Kestrel (cont.)

Name dog participant prevent.

Multi Android module project — using dagger2

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Vijay Chidambaram

Vijay Chidambaram

More from Medium

How the Census counts people living in group quarters and what it means for redistricting

Want to quantify solutions? Build a data center

⚔️ W A R ⚔️

The critically endagered Maui Dolphins need your support