Re: Another twin primes conjecture

From: David C. Ullrich (ullrich_at_math.okstate.edu)
Date: 06/19/04 

Date: Sat, 19 Jun 2004 13:07:39 -0500

On Sat, 19 Jun 2004 15:58:27 GMT, "Quinn Tyler Jackson"
<quinn-j@shaw.ca> wrote:

>David said:
>
>> I don't have any idea what you're talking about, regarding
>> mathematicians "taking risks". The point to pointing out that
>> James' prime counting algorithm is so slow compared to well-
>> known existing methods is that that is one big reason why
>> it's simply no big deal.
>
>I took a risk, and significantly sped up James' original posted algorithm
>using optimizations that have been around since Adam. Granted, I only
>reduced the invisible constant portion of O(n), and only slightly reduced it
>from 2n --> 2t to 2n --> ~1.7t. No big deal, indeed. Sure. And I only spent
>one day while flu-struck doing that. How could it be improved further? It
>might take a few risks to do it.
>
>[...]
>
>Sure, none of this is news to you, but it is to me. I could likely come up
>with stuff that's news to you, but in a different branch of maths than all
>this. I want to take a few risks to see what happens. The whole thing might
>be made faster as result. May not.
>
>The first sort most people learn is Bubble sort. Because it helps people
>grok the concept. Then they are introduced to qsort. James counter may be a
>bubble sort -- but it surely helps me grok the concept much better than the
>stream of probabilistic successive heuristics did.

Fine. When you appeared here it seemed you were wondering why
people around here took the attitude towards James that they
do, chastising us, explaining we should just dispassionately
discuss the math, etc. I was trying to explain part of one
answer to that question. Possibly you've figured out the
answer to that question by now. In case you haven't,
Bubble sort ("BS" below) gives us a hook:

Suppose that I appeared on a programming group and said what
you just said, that BS was a good thing because it helped
people understand what's what. People might agree, they
might disagree, they might yawn, no big deal.

Now suppose that I appeared on a programming group and
announced that I'd invented a new sorting algorithm that
was orders of magnitude better than anything that had
ever been done before. People examine my routine, and
they see that although I haven't literally copied the
source code out of any book they can name, it's clear
that I'm just doing a BS.

And then I spend months explaining that my BS is _not_
just BS, it's brand new. The people who are saying
it's just BS are evil - they don't care about Truth,
they're just trying to cover their ass because they're
embarrased they can't do as well. I claim that my
BS is much faster than qsort.

I _continue_ to claim that my BS is much faster
than qsort for a _long_ time, simply ignoring the
timings that the evil ones are posting as well as the
theoretical O(whatever) explanations.

Then I finally give up on that. After spending a long
time insisting that I've invented BS and that it's much
faster than qsort, I announce that the speed of a
sorting algorithm is not important. I continue to
rant about how _I_ invented BS, and the people
pointing out that it's been known for a long time
are just lying. Desperate for another reason
why my BS should be hugely important, having finally
been convinced that it's slow, I announce without
any explanation whatever that studying BS will probably
allow people to solve the P = NP question. And no
matter how many times people point out that I
haven't given any indication of a connection between
the two, I _continue_ to insist that my BS will
probably lead to a solution to P = NP.

After I spend about 8 or 9 years making posts
like that, often starting three or four threads
every day for a few weeks, I might eventually
become somewhat unpopular with the people who
I continued to call evil.

************************

David C. Ullrich

Relevant Pages

  • Re: Selection sort and bubble sort
    ... quicksort as a library function for the qsort() interface. ... I think the main issue is how one chooses the median. ... Falling back to bubble sort will not solve the problem. ... Radix and qsortinterface). ...
    (comp.programming)
  • Re: Template Metaprogramming in Forth?
    ... The QSORT is totally generic, the equivalent of a C++ template. ... The amazing aspect of the bubble sort example I ... mention is that the compiler transforms the general bubble sort ...
    (comp.lang.forth)
  • Re: Template Metaprogramming in Forth?
    ... The QSORT is totally generic, the equivalent of a C++ template. ... The amazing aspect of the bubble sort example I ... mention is that the compiler transforms the general bubble sort ...
    (comp.lang.forth)
  • Re: Selection sort and bubble sort
    ... >>> quicksort as a library function for the qsort() interface. ... I think the main issue is how one chooses the median. ... yes, but note that heapsort is not the simplest of sorting methods (unlike selection sort, or a modified bubble sort), which is why I had not done so before... ...
    (comp.programming)
  • Re: [possibly OT] the comparison function in quicksort
    ... desi_coder@yahoo.com (Debashish Chakravarty) writes: ... You mean qsort(). ... > information that might be required to compare the two objects, ... extension libraries have such a mechanism for this very reason. ...
    (comp.lang.c)
Quantcast