It's for fields where thoughts are evergreen (basically, anything to do with being human, literature, poetry, philosophy, etc) where this advice matters.
A good modern book on math, or chemistry, or compiler construction has more knowledge than any old one.
A good modern poet is not better than Shakespeare or Homer (and in many eras the poets are way worse than previous eras).
I'd argue that exactly because so much of it is cumulative, a lot of old technology texts stand up just fine when describing things like algorithms or a sub-field up to a certain level. The original paper on quicksort for example is just fine as an introduction to quicksort.
The books and papers that date are the ones that seek to tell you the best way doing something broad. An old text on the best way to sort in general will be date where descriptions of specific algorithms haven't.
Tell that to Claude Shannon, Fred Brooks, and Ken Thompson. "A symbolic analysis of relay and switching circuits", "The Mythical Man-month" and "On Trusting Trust" have aged fine, I promise.
- aged extremely well in some ways (it still takes 9 months to make a baby no matter how many women are assigned to the project)
- aged poorly in others (disk space is no longer an issue when deciding whether to comment on code)
- remained ahead of its time in others (its a good idea to have an architect to ensure the conceptual integrity of a complex system, rather than to have developers hack it out a bit at a time in a series of scrums)
Thought note that even those are barely ancient, they are at best a century old -- Lewis was talking of Plato as an example and in general centuries old classics, not whether someone should read Zola or Hesse.
Today very few would suggest reading Newton to learn physics in university (e.g. use it as a textbook). At best they'd tell to to read Newton to see how the thinking went behind early discoveries. But people use Plato or Shakespeare or tons of other centuries old writers as their core textbook all the time in philosophy and literature departments.
0: https://en.wikipedia.org/wiki/Hackers:_Heroes_of_the_Compute...
1: http://digital.library.upenn.edu/webbin/gutbook/lookup?num=7...
That's history.
Mathematics textbooks are just as likely as anything else to suffer from modern pedagogical theories that have not been tested by time and will come to be regarded as mistakes.
Not only do you learn geometry, but you participate in the same understanding of geometry that all later mathematicians started from.
I've always encouraged reading Euclid, Newton, Einstein. In my humble opinion, mathematics is much easier to understand historically, as it developed, and the best historical perspective comes from primary sources.
I must acknowledge, however, that for whatever reason very few people share my perspective on this.
It's the one with the most historical importance, but it hardly covers modern geometry.
I would argue that the fact that it does not cover modern geometry is exactly what makes it valuable. Learning is best as a process of rediscovery.