In some areas the list is certainly not accurate. E.g. in computational linguistics:

- Realization of Natural-Language Interfaces Using Lazy Functional Programming, Frost, 2006. I never ever heard of this article, with only 17 citations overall (in 5 years) it can hardly be considered important.

- In the entry of Transformation-based error-driven learning and natural language processing, Brill, 1995 (which is an important publication) it is stated that it "Describes a now commonly-used POS tagger based on transformation-based learning." Which is not true, since nearly everyone uses HMM, maxent, or SVM taggers these days because they give far higher accuracies.

Although it is far from perfect, the number of citations is probably one of the best manners to count importance. Someone actually did this per year for ACL conferences:

http://www.phontron.com/blog/?p=29

Obviously, there are other conferences, journals, etc. But it gives a pretty good overview of papers that are recommended. Also, there's the ACL top-10 rankings:

http://clair.si.umich.edu/clair/anthology/rankings.cgi

I took a look at the area that I am familiar with (database). If you'd like to gain more understanding about the area, it's probably better to look at required reading list from Berkeley and Stanford. (Note that the Berkeley list is longer.)

http://www.eecs.berkeley.edu/GradAffairs/CS/Prelims/db.html

http://infolab.stanford.edu/db_pages/infoquallist.html

Striking how old most of these papers are, many from the 60s and 70s, very few from the 00s.

Those who don't learn from the past are destined to reinvent it, poorly.

Some of the criteria are biased towards older papers. Influence is extremely biased-- it's rare for a paper to change the world immediately. Future influential papers written today may not actually earn the label for another 10-20 years. Breakthroughs and introductions are somewhat biased in general, but especially biased in computer science given the youth of the field and the fact that access to computers was rare and difficult before the 60s.

"Latest and greatest" is not biased, in theory, but in practice it can be a lot harder to identify what truly is the latest and greatest as opposed to what is merely the most popular.

Somebody once said "If you want to learn something new, read an old paper".

If you're an aspiring computer science researcher looking at these articles, I cannot recommend enough Hamming's speach "You and Your Research." (http://www.cs.virginia.edu/~robins/YouAndYourResearch.html)

He talks about the little optimizations you can do in your life to take your research to world-class level.