Talk:Perl OpenGL

From Wikipedia, the free encyclopedia

[edit] Neutrality

I think that the neutrality of the performance section is disputed, for two reasons:

  • all benchmarks are coming from the website distributing the software (graphcomp)
  • all the edits (except the last ones, made by me to add wiki links) concerning this part were made by one contributor, who also owns the same company distributing the software

Claims in this part should be alleviated, or referenced against other sources than only the main company's source. Hervegirod 14:38, 9 June 2007 (UTC)

[edit] Response

  • All benchmarks include source, and are freely replicatable.
  • The POGL vs SDL::OpenGL vs PyOpenGL benchmarks were based on a 3rd party OpenGL benchmark called Trislam, written by Geoff Broadwell; these benchmarks have been independently reproduced by him.
  • The Perl vs Python benchmarks include comments from PyOpenGL's author confirming why Python would be slower than Perl.
  • These results have been repeatedly headlined on OpenGL.org - generating no claims to the contrary.

The fact that the same author created the libraries, ported the benchmarks, and authored this article, should not be taken as an indication that the data is wrong.

Unless it can be demonstrated that these benchmarks are incorrect, these neutrality notices should be removed; the author has provided all means for readers to verify the claims. Grafman 15:30, 20 June 2007 (UTC)

[edit] Resolution

Per recommendations on the Village Pump Policy page (Neutrality Resolution topic), I am removing these neutrality tags. Grafman

Please stop creating new sections for each new response, it's unnecessary. I don't see any actual discussion on this page beyond your one reply, which is not grounds for you to remove the tag. I completely agree that these sections are a violation of NPOV policy. I will be re-adding the tags and removing sections which aren't even relevant to the article (as outlined below, it's not relevant if it's not regarding a benchmark specifically comparing OpenGL performance in different languages).
You are also violating No Original Research and Conflict of Interest policy. This is research conducted by the authors of the software themselves and is added by them. There is no alternate research on this or academic/scientific critique (positive or negative), there are no independent researchers confirming the validity or lack thereof of these benchmarks, they aren't in any sort of academic/scientific or otherwise notable publication, and everything from the benchmarks is being stated as fact (blatant violation of NPOV policy--it doesn't matter how good they are, either) so they are totally inappropriate for Wikipedia. The writing of it is also clearly slanted to promote Perl OpenGL and Perl in general. I shall emphasize that it doesn't matter if this is the best benchmark and library in the world, that is still a clear violation of Wikipedia policy.
The problems with your argument are as follows: 1) it doesn't matter to Wikipedia if the source is available, 2) utilizing some other benchmark software that isn't recognized by anyone as a standard is irrelevant, 3) your comments from the Python author are about a deprecated version of PyOpenGL, 4) this doesn't apply to other benchmarks, 5) blogs are not generally considered reliable sources and being on OpenGL's blog is hardly equivalent to being published, and 6) the lack of claims to the contrary (aside from the ones here and in other places) doesn't actually make it any more suitable for Wikipedia no more valid, it just means that no one has bothered to write contrary benchmarks, that you know of.
There are further problems. You include comparisons based on generalized and unrelated benchmarks from two sources regarding Ruby and Java--those don't even apply to the article. The only ones that apply need to compare the usage of Perl OpenGL to other OpenGL libraries. PERIOD.
Other problems (NOTE: it violates policy even without these--I just wanted to be thorough): 1) the benchmarks and their ports, including the Python one, are written by obviously biased parties (including the Perl OpenGL authors themselves), 2) these are simplistic tests and don't really cover the large potential range of functionality (real tests for 3D software are pretty elaborate and take years to develop and even then are flawed), 3) the ports are designed to mirror the Perl code as close as possible, line for line, even though this excludes the possibility of optimizing it well for the language being ported to, 4) this is actually a comparison of library implementations, as Perl itself and the libraries being used are written in C--so clearly, the C code is just poorly optimized (anything suggesting that interpreted Perl is faster than C is a joke), 5) the comparisons extend beyond the benchmark, into functionality not even tested like string process (and again, this is just a library implementation issue). -Nathan J. Yoder (talk) 13:01, 5 May 2008 (UTC)