Thursday, March 5, 2009

research dilemmas

I'm blocked.

Most of the people in my group are furiously trying to finish a paper for SOSP, and I'm off on my own thinking about web security.

It's not that I don't think this is an important problem -- I truly believe that the security model in today's browsers is a patchwork of brokenness, and could perhaps benefit from being rethought from the ground up. However every time I try to do that I end up meandering off into nothingness, or reading a blog. And it doesn't help that it seems like the problems are getting solved as I meander -- the Gazelle browser, from Microsoft, seems like a step in the right direction.

Also, I'm taking two classes this semester, Natural Language Processing and UI Design, which seem to be taking up increasing amounts of time. And of course there's my other job. It's so much easier to hack on Native Client than try to come up with a viable research project. By the way, we're having a security contest. If you're interested in security, or just interested in that sort of mechanism for validating your self-worth, go try to find some bugs!

I'm thinking of switching. To multicore, which the rest of PDOS seems to be working on, or something scalable-webapp related. I'm not sure how this research thing is supposed to work -- are you supposed to keep beating your head against a wall until something happens, or is it sometimes appropriate to give up and move on? How do you tell? Thoughts are appreciated.


  1. Hi Neha,

    It's definitely sometimes appropriate to give up and move on, and it's very difficult for inexperienced researchers to tell when that point is. That's where advisors come in handy.

    When my advisor asked me to make a thesis outline, I gave him 3 alternate ones. He told me if I did everything in those 3 outlines it would take me 10 years and I'd have tenure. He picked the one where I was furthest along, and a year and a half later I had my Ph.D. An experienced advisor has a good sense of what's achievable in how long, and what's too ambitious.

    The good thing about a thesis outline is that it breaks up a huge looming project into smaller tasks so it's easier to see one is making progress. On the other hand, in my last semester of graduate school I had gained momentum and came up with ideas that I was really excited about, that weren't in the original plan.

    My postdoc mentor likes to talk about research in terms of asset allocation. You invest some of your time in what you're sure you can do, like investing in bonds with steady returns. But you also invest in the ideas you're really excited about that can have a high impact, even though you're less certain of how to proceed, like investing in stocks with great potential for growth. (Of course, this was all before the current market situation, but you get the analogy.)

    Would switching to scalable web-apps involve working with a different advisor?

  2. thanks for writing, Ruchira, this was really helpful.

    I'm switching what I'm focusing on. I feel better about the new thing. Switching doesn't involve changing advisors :) My advisor is happy to advise me on whatever I want to do -- that's the problem! I'm not research-mature enough to come up with well-defined problems that have a high likelihood of being successful.

    An outline might help. I get trapped in this phase where I'm constantly exploring the area and I don't do anything specific.

  3. hey neha just read your blog day before and I wanted to share this link-
    i am just an amateur undergrad in computer science and you know dont have much idea about your academics but this video really raised my spirits..u know so if u are still down do watch it..

  4. thanks dupree! I heard about that but haven't seen it, though in fact, I just checked out his book from the library. It seems like he was a very inspiring professor.

  5. Neha, whatever you do, don't do it for the sake of getting a few more SOSP/OSDI's accepted. It'd be unfortunate if this is what a new grad student thinks is important compared to actually sitting and doing good research! If you look around, this is what happens in most CS circles.