Wikipedia talk:WikiProject Computing/Archive 13

This is an archive of past discussions on Wikipedia:WikiProject Computing. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Archive 10

Archive 11

Archive 12

Archive 13

Archive 14

Power Architecture on Wikipedia

Power Architecture is a marketing term that was promoted by IBM and Power.org members during the mid- and late-2000s. It is an umbrella term that was used to describe anything that had some relation to the POWER, PowerPC, or Power ISA architectures. It has been out of use by IBM et al. since IBM founded the OpenPOWER Foundation and Power.org became defunct sometime during the early 2010s.

Wikipedia has followed IBM et al.'s usage conventions of this term, and has applied it liberally throughout the encyclopedia, which has resulted in:

An article: Power Architecture
A navigation template: Template:Power Architecture
A category: Category:Power Architecture
Inclusion in the text of up to 236 articles (this is based a in-text search of "Power Architecture")

There are two problems with this term. Firstly, it's technobabble. It sounds like as if it's a well-defined technical concept, but it isn't. Secondly, it has a large potential to confuse. "Architecture" is an too-overloaded term in ordinary English and in computing. In computing, it could refer to a multitude of different things depending on the context. For example, in the context of hardware, "architecture" could refer to the architecture of a computer system, the architecture of a computer subsystem (processor, memory, I/O, etc.), or the abstract model of a computer (also called computer architecture or instruction set architecture).

My concern is what do readers think of when they encounter "Power Architecture"? Besides the ordinary uses of "Power" in English, IBM has used "Power" or "POWER" in a number of different contexts. Originally, it was an instruction set architecture (IBM POWER instruction set architecture). Then it was a series of processors IBM POWER microprocessors. Most recently, there's a family of computers called IBM Power Systems. Then there are nouns that include "Power" as a part, such as PowerPC. While Wikipedia can't do much about the names of these things, since it must call something by its name, it can omit "Power Architecture" when it isn't relevant to the article, doesn't explain or clarify anything in the article, or doesn't improve the article.

To demonstrate that this is not theoretical, consider the Power Architecture article. Its purpose is to explain what the Power Architecture is, yet it's confused. The article provides a glossary to define the various terms that could be confused with each other, and then proceeds to make two glaring self-contradictions:

It has a large section about the Power ISA, which the article earlier defined as something that is distinct from the Power Architecture (the former is a computer architecture, the latter is a trademarked umbrella term that may refer to things with a relatation to the former).
It concludes with a long confused list of things that "implement" the Power Architecture, which is a misuse of "implement", and a misunderstanding of the Power Architecture. If Power Architecture is an umbrella term for various related things, then it cannot be something that can be implemented—it's not a specification.

As another example, consider Template:Power Architecture. The way it is structured is confused. Power Architecture isn't a standard which is "made" by various manufacturers. The standards are the POWER, PowerPC, and Power ISA architectures. Yet there is no consistency in the way that template is organized. That cannot be conducive to easy navigation.

The importance of this term is also overstated on Wikipedia. Category:Power Architecture duly follows Power Architecture's definition; every article that meets the definition is categorized in that category. However, the majority of the articles are about topics that predate the introduction of Power Architecture. Given the nature of computing, the majority of the sources for these articles would likely predate the introduction of Power Architecture as well. Consequently, most, if not all, sources that these articles cite, would not support the claim that these are Power Architecture topics. To say that they are Power Architecture topic places undue weight on the marketing from IBM et al.

Because Power Architecture is a vague, ill-defined marketing term and has a large potential to confuse, what is its encyclopedic usefulness? I would argue it has none. It's not a definition that explains and clarifies, and it doesn't group topics together better than the existing schemes that are founded on technical criteria. It exists on Wikipedia because it does elsewhere; and its conventions are followed because Wikipedia can, not because Wikipedia should.

The organization of the topics that fall under the Power Architecture umbrella should be determined by what makes good sense for an encyclopedia. To this end, I think:

The Power ISA content in Power Architecture should be split out into Power ISA.
The content in Power Architecture about the term should be merged into Power.org. It is sufficient to explain that Power Architecture is a marketing term in the article about the organization which spawned it.
Every mention of "Power Architecture" in contexts other than as a marketing term should be removed from Wikipedia. There's no need to insert this term into every topic that Power Architecture deems falls under its umbrella. Where this term has been used mistakenly instead of the correct term, the mistake should be corrected. For example, if an article about a processor describes it as a Power Architecture processor, then the actual architecture the processor implements should be mentioned instead.
Category:Power Architecture should be deleted. There's no need to categorize articles under it just because IBM et al. say their topics retroactively fall under the Power Architecture umbrella.

99Electrons (talk) 00:09, 22 February 2019 (UTC)

Most of this sounds reasonable to me. The only standard other than the Power ISA that I see mentioned on Power Architecture is the Power Architecture Platform Reference; other standards can probably have their own pages as well.

Power.org appears to be dead; is that a further sign that "Power Architecture", as a concept, is also dead? The OpenPOWER Foundation page (which I just fixed to use https: for all openpowerfoundation.org links, as that site seems to reject Boring Old No-TLS HTTP) says that "Power.org is still the governing body around the Power Architecture instruction set" - is that now out of date?

And is there anything in the PowerPC ISA that isn't in the Power ISA? If not, the two ISAs should perhaps be discussed on the same page - perhaps have a single page for the ISA, have "PowerPC ISA" redirect to it, and combine ISA information from PowerPC and Power Architecture on that page. Guy Harris (talk) 00:47, 22 February 2019 (UTC)

Curses to IBMs terrible naming strategies. And I agree with Guy Harris.Jacona (talk) 14:44, 22 February 2019 (UTC)

@Guy Harris: I'm almost certain "Power Architecture" as a brand is dead. I can't find any non-incidental or legacy use of it on IBM's website (which seems to use the OpenPOWER brand instead). It's the same for the OpenPOWER Foundation. I can't find any current use of Power Architecture using that website's search function or with Google.

I think merging the PowerPC article and Power ISA content from Power Architecture is a better solution than having a separate article for the Power ISA. The latter is clearly an outgrowth of the former: The manual for Power ISA 2.07B ([1]) says in the preface that Power ISA 2.03 was created in 2006 by combining PowerPC 2.02 with PowerPC Book E, which was then not a core part of the architecture. Note that in Power ISA 3.0 B, all of the embedded features introduced in 2.03 were removed (as documented in the preface of [2]). I personally don't think this is relevant to the question of whether PowerPC and Power ISA can be covered in the same article, as 3.0 B is still a development of something that began with PowerPC. 99Electrons (talk) 21:39, 22 February 2019 (UTC)

Comment: I forgot to mention this in my OP, but there's a few discussions concerning "Power Architecture" and "Power ISA" in category names that has some relevance to this discussion:

99Electrons (talk) 00:51, 23 February 2019 (UTC)

My 2¢ is that most of this confusion is due to IBM's inconsistent use and definition of what's what. I've tried my best to unclutter it, but hell, it's not easy. I cant be as invested in Wikipedia as I've been in the past, so I gladly pass on the enthusiasm to whomever what to keep on pushing. Most suggestion seems reasonable. You have my blessing going forward. -- Henriok (talk) 10:47, 23 February 2019 (UTC)

Agree The phrase "Power Architecture" is really dumb and ambiguous, to distinguish its usage for several different things, there is need to use correct phrase. Editor-1 (talk) 13:26, 6 March 2019 (UTC)

Discussion of LWN.net on the reliable sources noticeboard

There is a discussion on the reliability of LWN.net (formerly Linux Weekly News) at the reliable sources noticeboard. If you are interested, please participate at WP:RSN § LWN.net for Draft:NumWorks. Thanks! — Newslinger talk 02:45, 7 March 2019 (UTC)