graduate school - How does one keep herself updated with new research without forgetting older results?

The title is fairly generic in nature, so I'm trying to elaborate in the body. I'm interested in answers pertaining to Theoretical Computer Science (TCS), but I'm certain the question would be equally relevant in fields which have been around for more than a couple of decades, and hence I hope to get responses from researchers in other disciplines as well!

What I'm trying to understand is how researchers (who have been working for quite sometime) in any specialized field (like TCS) keep track of results that have already been published - not just seminal results, but also results which have a lesser (but not insignificant) impact on the field but were published years or decades before. At the same time, one has to keep track of results being published in (at least) the notable conferences in the current year as well, in order to absorb the new ideas presented there and incorporate/extend them in one's own work.

I find it kind of incredible to believe that all of the above is possible without any kind of disciplined approach to reading and subsequent assimilation of the ideas on a regular basis - which is why I'm asking members of academia about their experiences/practices on this. In particular:

1. How frequently (if at all) do you revisit "classic" results?


2. Do you keep written/electronic notes on a regular basis to keep track of continuing progress in a field (say, for instance, inapproximability results for geometric problems) - or do you prefer to keep it all in your mind?


3. To keep track of current state of art, do you only attend/read Tier-I conferences, or do you get useful ideas from results published in Tier-II/III conferences as well?


4. What kind of time/resources would you typically set aside for reading, as opposed to working on a problem?

In short, I'm trying to find what kind of things would you expect a top researcher to "know" off the top of his head, and at what level of depth - and how would you go trying to maintain that level of perception over the years?

(I understand that the question IS subjective, but I'm hoping that it satisfies the guidelines for a "good subjective" question!)

Answer

Here are a few additions to Suresh's list (converted from a comment at TCSgrad's suggestion):

• 
  

  Don't even try to remember details. Just remember that somebody published something related, and use Google (or Mendeley or Papers) to find the paper again when you need it.

  
  


• 
  

  Chase references. Interesting papers tend to cite other interesting papers. When you read any paper, also look at the papers in its bibliography.

  
  



• 
  

  Chase citations. Interesting papers tend to be cited by other interesting papers. When you read the paper, find other papers that cite it (via Google Scholar, for example) and look at them, too.

  
  


• 
  

  Follow whims. If you see a paper with an interesting word in the title, or an interesting figure on the first page, at least read the abstract.

  
  


• 
  

  Don't try to read everything. That's impossible. Just try to read a little more.

  
  


• 
  
  

  Stop reading. Eventually you have to do your own research. Don't worry about reinventing the wheel; sometimes the best (and even fastest) way to understand what someone else did is to ignore them, figure it out yourself, and then read their paper.