Talk:GotoBLAS

Proposed merge with OpenBLAS

I'm not too convinced of the separate notability of OpenBLAS, which is why I initially wrote a single page about both GotoBLAS and OpenBLAS. The papers about it have relatively low citation scores, and what else I could find online are blog posts [1] and manuals of other software packages [2], not in-depth coverage of the OpenBLAS product. QVVERTYVS (hm?) 08:11, 16 May 2015 (UTC)

Well I originally moved it to OpenBLAS since I figured Wikipedia ought to emphasize what is relevant nowadays instead of what was relevant 10 years ago. After all, OpenBLAS is at least as notable as GotoBLAS, if not more (e.g. 41k vs 18k results on Google). But the move was disputed because GotoBLAS still has some historical significance.

Okay, that's reasonable since ultimately they are separate projects maintained a completely different set of developers, so they should really be on separate articles. The fact that one is a fork of the other isn't important anymore since codebase has diverged several years ago. Furthermore, end-users seeking support for either software should use their respective communication media (e.g. for OpenBLAS there is a mailing list).

Given that it is a library for end-users, not the kind of software intended purely for making research, I would not expect citations to be a good measure of its notability. I wouldn't be surprised if some users aren't even aware they are using it in their research, since it's a rather low-level package that drives other major packages like Matlab or Numpy.

It doesn't have a very detailed manual other than the Wiki as mentioned below, which is unfortunate but that's the norm for BLAS-like packages. Most such packages expect you to just look up the documentation for the BLAS API. Most such packages don't even have the luxury of a Wiki. (AFAIK the only decent source of docs for BLAS is MKL and maybe Netlib BLAS if you count the source code comments as such.)

Here are some random facts I can gather about OpenBLAS:

• It has a registered domain: http://www.openblas.net
• It is actively maintained by at least 3 developers: Zhang Xianyi, Wang Qian, Werner Saar; last commit was 3 days ago.
• It has 536 stars and 147 forks on GitHub.
• It has a Wiki with FAQ, Installation Guide, etc.
• The packages are maintained in Debian and Ubuntu repositories.

So I would argue it's not exactly an insignificant project. In fact I would consider it to be quite notable by the rather low bar set by typical numerical software that isn't Numpy-popular. --Fylwind (talk) 09:26, 17 May 2015 (UTC)

Wikipedia emphasizes what has significant coverage in reliable third-party sources. GotoBLAS has that kind of coverage because in its day, it was the fastest non-vendor BLAS, its implementation shattered some myths about how to implement linear algebra and because of its rather peculiar history. None of the sources you found really belong to the category of sources that warrant a separate article (I had found pretty much the same ones):

• Google hits by themselves don't establish notability. Similarly, GitHub stars and forks are not significant coverage.
• Anyone can claim a domain name, set up mailing list, set up a GitHub wiki, write documentation for their software project.
• Actively maintained doesn't mean notable. Abandoned by its developers doesn't mean non-notable.
• Inclusion in Linux distros is not significant coverage either. It establishes existence, not notability. The sl command is listed at the disambiguation page SL, but it doesn't have its own article.

So what we have to go on are the articles written by the authors, which are affiliated primary sources.

I'm not saying OpenBLAS is insignificant (I found it important enough to write about in the first place), but I am arguing that it doesn't meet the bar for a separate article. QVVERTYVS (hm?) 11:16, 17 May 2015 (UTC)

To me, a separate article is justified. For Windows users it is very relevant as this seems to be the only hassle-free download-and-use optimized BLAS for use with Visual Studio without having to resort to Cygwin and the like. Just my 2 cents.153.96.12.26 (talk) 15:17, 2 November 2015 (UTC)

Wikipedia is not a manual or a software suggestion service. QVVERTYVS (hm?) 16:00, 2 November 2015 (UTC)

Closing; stale proposal with no support from people other than the proposer, and opposition, despite more than 2 years. Klbrain (talk) 22:22, 21 September 2017 (UTC)