Talk:API

some sections require cleanups

I noticed some types along the way as well as lots of mixups between API and binary compatibility and technical glitches in the "Implementations" section. Specifically, what seems to be referred to as a "compatible API" is really called "binary compatibility" (and occasionally i've seen it referred to as an ABI or ABI level) and is not necessarily related to APIs. also the paragraph dissing apple about "emulation modes" while praising microsoft for "backwards compatibility" should probably be removed; it's neither relevant nor technically correct. the emulation mode is slow because typically it's emulating a completely different CPU architecture, something that cannot be done in any non-slow way. speaking of which, "the windows api" is available on several distinct CPU architectures, most of which are not binary compatible with each other, meaning that programmes written for one architecture will need to be recompiled to work on another, regardless of whether they were programmed solely towards the windows api alone or not. --82.113.121.194 (talk) 23:33, 22 January 2010 (UTC)

older entries

Sunit Pinto is working on API.

Wait a minute, SNMP isn't an API is it? It is a protocol. Don't know much about CORBA, but I wouldn't call DCOM an API. Robneild 15:53, 16 Feb 2004 (UTC)

The definition at the top of this document is too inclusive, with the result that protocols would be considered APIs. The common use of the term is for a collection of classes or functions. CORBA is many things (protocol, language mappings, type system...) and it happens to include an API as well - the ORB API. But most people would not call CORBA an API. At most, it is a language for expressing APIs. Yaronf 23:31, Feb 18, 2004 (UTC)

Your definition of API's is way too inclusive. The API is the programming language interface provided to users, in that programming language. This is different than protocols, which usually represent the "wire" format of the actual communication. This might be the may web interfaces, IIOP in CORBA, or the conventions used in a socket call. Representing these in XML and ontologies are the next evoloution of Protocols - but not of API's. I contend that API's are to be minimized as they create basically the next N squared problem in that to get them implemented in every language, you have a very uphill battle, let alone making changes. Protocols, on the other hand, can evolve and you let the languages find their way to create the protocol. The reference to ontologies would fit with the Protocols, but not with the API.s

"It is generally lawful to analyze API implementations in order to produce a compatible one" ummm. If the wine group looked at a Microsoft WIN32 implementation they would be in trouble. I thought they would steer clear of Microsoft's implementation. Guess it depends what is meant by analyze. I would have thought once you have a specification for the API that is all you need e.g. "int time = GetTime()". Robneild 14:44, 17 Oct 2004 (UTC)

Too technical

This article is much too technical for even a computer-oriented person. It needs to have the introduction completely re-written and the technical details relegated to a lower section. Of course, I don't have the time...195.176.162.18 (talk) 07:00, 2 September 2008 (UTC)

API Category

I was thinking about creating Category:Application programming interface, creating a reposity for all the API's that exist in wikipedia, (DirectX, win32, opengl, etc)

• I like this idea Jontce 14:53, 15 November 2005 (UTC)
• I couldn't find a better section and I'm new to commenting, so I've included my comment here. Let me apologize in advance if my thoughts are not in the proper area. I wanted to comment on the over-explanation of ABI's in this article. Someone needs to go back and clean that up. If ABI plays an important role in explanation, then so be it, but it needs to just have a link to another wikipedia section explaining ABI. For someone who does not understand all the basic concepts behind API or ABI, the additional information of ABI here simply threw me off. —Preceding unsigned comment added by Chrisban35 (talk • contribs) 13:52, 29 March 2009 (UTC)

Little bit of confusion...

This may sound stupid but what is the difference between an API and a library?

The API is only the interface. —Preceding unsigned comment added by 80.38.67.99 (talk) 09:06, 16 June 2008 (UTC)

---

Simply, an API is a contract to adhere to while a library is an implementation of that contract.

In my experience the term 'API' is used interchangeably to mean the library, its interface and the 'contract'. I'm thinking especially about the Java API. --Uncle Ed (talk) 23:25, 7 January 2009 (UTC)

That still doesn't make sense. I think what you guys are really talking about, is the table of symbolic labels (sometimes including expandable macroes) that the symbolic assembler pulls in from disk, processes, and generates executable object code. Good assemblers let you import other people's tables of labels, and mix them with other people's source code, and more or less bypass the first pass of assembly before generating code. If that is what you are talking about, the concept of an API is completely illusory. The execution context really depends on which microprocessor you are working with, in particular, and how the lines of I/O are being dealt with. For instance, there are microprocessors with extremely limited stacks, and I doubt that a single one of your so-called API's would work with them. Dexter Nextnumber (talk) 04:09, 21 December 2009 (UTC)

Framework,API,Library,Protocol

Dear All

       I think it is gona be a big miss understanding here, as a

programmer in the first place and Systems Analyst, i can tell that we are talking about this Pyramid:

                 -----------Program-----------
              -------------Libraries-------------
          -------------------APIs-------------------
   -----------------------Framework-----------------------

---

Operating System Layer--->Protocols---> Operating System

the programmers usually make their own libraries using calls to APIs which usually be a part of a particular framework which must be based in a particular operatin system, and the operating system uses a protocol to manage the communication with another operating system.

as an example, if we are talking about :

---

The Program------------------------------

Programmer Made Library to manage what ever tasks---- .NET APIs (System.Windows, System.Windows.Form...etc)

---

.NET Framework-----------------------------

---

WINDOWS XP------------------------------

Best Regards Hossam Abbas

---

That pyramid is deceptive. It is possible (for example) for a single API to be spread across multiple libraries - or for one library to implement multiple API's. An API is a specification for an interface - typically between a program and a library - but possibly between two libraries - or between a 'plugin' and a program. An API is more like a piece of documentation than an actual library. A single API can apply to multiple implementations - and hence to multiple libraries. I can write an OpenGL application that links both to (say) nVidia's OpenGL implementation and the Mesa OpenGL implementation. Two implementations of the same API. The reason we need the concept of an API is precisely because the word library doesn't work.

SteveBaker 14:41, 2 February 2006 (UTC)

Too many links.

I don't think we need all of those links to Example APIs - we are getting dangerously close to link-spam. Also, the WP:MOS recommends only using external links for references (which ours are not) - or to cover things not adequately described in the article or other Wikipedia articles. Since we have a good list of internal links to example API's, I move to delete all of the external examples.

There are two external links in the example list of API's (MediaWiki and Drupal). I suggest we replace them with internal links to the project pages, because external links should be placed under External links section. The external Drupal API link already exists at Drupal, and the external MediaWiki API link could be moved to MediaWiki. --Luen (talk) 09:45, 5 December 2008 (UTC)

My feeling on the matter is that if an API is not important enough to have its own page, independent of the product it is an API to, then we probably shouldn't be linking to it. So I'd just delete these two links, along with everything else in the rightmost column, and the iPhone API. JulesH (talk) 16:19, 6 December 2008 (UTC)

This article needs to get it's Featured Article status back again. SteveBaker 22:21, 31 March 2006 (UTC)

Application Programming Interface OR Application Program Interface

Just to check; technically is it an Application Programming Interface or is it an Application Program Interface? I know this is very picky, but an Encyclopedia should get it right. (please don't flame me). Thanks --Skoorb 17:26, 1 April 2006 (UTC)

I believe it is Application Program Interface since the interface remains between the application and the library long after the act of Programming has been completed. One might even (theoretically) define the API after all of the programming has been completed - for example when you decide that a part of an existing application could be split off and made into a library if only you wrote down a specification for the API. However, it's a small thing - and since this is one of those informal terms that has gradually taken hold, it's perfectly possible that some users take one meaning and others the reverse position. SteveBaker 17:34, 1 April 2006 (UTC)

I have seen API defined as an Abstract Programming Interface. ChopMonkey 15:08, 2 October 2006 (UTC)

Another term which is often used and makes IMHO most sense is 'Application Programmer Interface', as it is an interface for the developer to write a new application.

I agree with the comment above, 'Application Programmer's Interface' is also used often. 88.255.168.161 08:17, 26 June 2007 (UTC)

"An API is to the programmer what a GUI is to the end-user. The 'P' in API stands for "Programmer", not "Program", to highlight the fact that APIs are used by programmers, who are humans." Maybe there should be a reference to the fact, that API has different decronyms which lead to different interpretations. 85Pando (talk) 11:34, 7 May 2010 (UTC)

API redirects here

Edcolins Why would you restore this. API doesn't redirect to this page it goes to the disambiguation page. Try it. jbolden1517^Talk 18:26, 5 May 2006 (UTC)

API should redirect to disambig not here. It's totally unfair to other disciplines to monopolize just because it's common in our field. --Treekids 19:25, 1 October 2007 (UTC)

Pronunciation

So how do most people pronounce API? A P I, or appie?

I—and every programmer I've worked with—pronounce it "A" and in "day", "P" as in "Pea in a pod", and "I" as in "I love you." HTH — Frecklefoot | Talk 14:30, 29 August 2006 (UTC)

"appie" is too easily confused with other words, many of them emotional. Programmers hate that ;-) --Treekids 19:31, 1 October 2007 (UTC)

Open versus closed

I would like to have a better description of when an API is "open" versus when it is "closed". The current definition focuses on rights of usage and access to source code, but I think the idea is broader than that. For example, I think an API is also "open" when it is clearly documented and able to be extended through some defined process. I also think standardization plays a part in this discussion. Thoughts? --Mwfnwa 13:00, 5 September 2006 (UTC)

---

I agree that this aspect needs work. the emphasis is on the financial cost of access and application. how about addressing the issue question of who gets access? as i understand it, an open api (or any open technology) is equally open to all comers.

in other words, if make my technology available free of charge to my friends, charge others for it, and do not make it available at all to those i regard as potential competitors, my technology (api or not) is not open. (i seem to recall that one of the standards bodies makes this an important part of their definition of "open standard." ie, not whether there is a charge, but whether or not the same thing is available to anyone willing to pay and at the same price.) - ef

Multi-language API's

Shouldn't we have an example of a language-independent API, such as DOM?:

The Document Object Model is a platform- and language-neutral interface that will allow programs and scripts to dynamically access and update the content, structure and style of documents. (From http://www.w3.org/DOM/)

Please, take not that there is also at least another internation meaning of "API": Active Pharmaceutical Ingredient. I don't know how to do, but I think that Wik should mention it. Ciao

Source code definition a requirement for being an API?

I don't believe it is true that an API may only be defined at a source code level. I don't see any reliable sources that use this definition (a google search that eliminates duplicates and obvious copies of this article turns up nothing useful). Furthermore, I see plenty of examples of APIs that include details that are not source-code specified, e.g., the Win32 API's reliance on Microsoft's structured exception handling implementation, the MS-DOS API (which specifies register sets that must be initialized to particular values and machine instructions that must be executed to perform a particular function - see, e.g. [1] for evidence that the term API is used to refer to this), and so on. The insistance on including the phrase "source code" in the definition, when the next sentence makes it clear that source code is the usual way of defining an API, is simply unsource POV insertion.

You are right that some vendor APIs include information that would normally be considered part of an ABI. I suspect that the intent on the vendors part is to produce a single document, rather than have separate API and ABI documents. This vendor usage ought to be mentioned in the article. In many ways the Microsoft APIs are heavily tied to Windows and might be called ABIs with an API veneer. I would also point to the P in API, it is a programming interface; one that gets used when writing software, ie, source code. Derek farn 16:25, 6 August 2007 (UTC)

I'm not going to remove it again yet, to give those who favour its inclusion a chance to justify it. But I think they'll have a hard job. JulesH 14:49, 6 August 2007 (UTC)

When source code is not provided, there is a real issue here that the API definitions are almost always at best ambiguous and underspecified. Simply, the black box mapping is never specified in practice in these cases, nor is the mapping unambiguously defined (basically in contrast to when the API is accompanied by source code and an open source build tool chain -- http://cm.bell-labs.com/who/ken/trust.html ). Hozelda (talk) 00:27, 30 January 2009 (UTC)

Decreasing quality

I note that this version of the article is in many ways superior to the current one. It's easier to understand, containing as it does a good example of an API described in terms that are relatively easy for a layman to understand (it could have been improved, but as it was it was simply deleted), and the text appears to me to be better written. Why have we destroyed a perfectly reasonable article? JulesH 14:55, 6 August 2007 (UTC)

Ditto! The current article is too technical. And I just saw that we actually have a category for that! Nice! Strawberry Island (talk) 21:56, 13 February 2008 (UTC)

Wow, the word "inculcate" is used in first paragraph of the introduction! Can't someone find a way to describe whatever is trying to be said there using simpler words? I know this isn't on the simplified English version of Wikipedia, but, come on, I am a native English speaker, I had to look the word up, and I STILL don't know what they were trying to say there. Most people who are viewing this article are NOT going to be a combination of a college English major and programmer.Mmortal03 (talk) 10:48, 25 February 2008 (UTC)

Disambiguation

Hihi

Working on cleaning up disambiguation pages...the one for API (disambiguation) has well over 500 links to it and more than 99% are meant for this page...I see there's been some discussion in the past about this...we are not saying that the programming language is the ONLY or even BEST definition, but its the one that people are most likely looking for when they type API into the search box. So I'd ask that you consider allowing me to do the redirect and put the hatnote on this page please? Thank you for your consideration. Legotech (talk) 17:49, 23 February 2008 (UTC)

requesting clarification of the term 'Remote Procedure Call'

Hello,

I would like to quote what was said about language independent APIs - " Language-independent APIs are written in a way that means they can be called from several programming languages. This is a desired feature for a service-style API which is not bound to a particular process or system and is available as a remote procedure call."

Now what this implies to a relative newcomer like me is that one can have APIs written which can be accessed by programs written in whatever language there is out there. But after reading this - http://en.wikipedia.org/wiki/Remote_procedure_call article on 'Remote Procedure Call' what comes to my mind is that 'remote procedure call' is a mechanism whereby the computer could have one of the sub-routines of the currently executing program to run in parallel on a physically remote computer. And it clearly states in the article that it is a paradigm of distributed computing. So can anyone please explain how does this relate to 'language-independent APIs' ? Some scenarios would be of help here.Aijazbaig1 (talk) 11:38, 15 July 2008 (UTC)

APIs

As a technical writer working on APIs, it seems to me that a bit of grammar help might be a good idea in this article; too often "API" is used to mean one call of many that make up an API and thus too much documentation uses "APIs" where one API (consisting of a multitude of API calls/functions) is meant. --iFaqeer (talk) 11:31, 17 July 2008 (UTC)

• You're right. Also, perhaps API's might be appropriate in places. Sounds like you are the ideal person to make the edits. Derek farn (talk) 13:29, 17 July 2008 (UTC)

API management (web services)

This section is almost by definition not about APIs, but how they fit into a larger sofware enterprise. I think it would be better off under Web Services or some such. There are all kinds of other troubles with this article (e.g. I think it is too Web-centric) but the new section, while in good faith, I think belongs elsewhere.

Perhaps put in a separate article and then link to it using main or something?

Best wishes.

SimonTrew (talk) 16:33, 20 April 2009 (UTC)

I'm of the opinion that we should probably purge this article of most (if not all) references to web service APIs. They are fundamentally different, and are better thought of as a protocol than an API, to my mind. Certainly they don't fit the description of the subject in the lead section. JulesH (talk) 16:54, 20 April 2009 (UTC)

I am roughly in agreement; I don't they are totally* irrelevant, but agree most of the time here they are being used to at a very broad description level. The google maps API, for example (I don't know it well) is to my mind an IPI: you submit some XML or SOAP or something to a well know endpoint (an URL of some kind, an http address I think) and it needs to have coords and scale and the other things you expect from maps. Or the google search API probably simpler and better. But "Google" is not an API (it's not suggesting it is here, just using it to try to make a useful distinguishing feature). So yeah most of it is way above the abstraction level of an API. SimonTrew (talk) 17:41, 20 April 2009 (UTC)

BTW I wondered whether COM should be thought of as an API, protocol (it's not really a protocol) or whatever, but on balance I think the amount of stuff defined in COM (and Java etc) qualifies as APIs if they are not, by a 100% strict definition, an API of themselves. SimonTrew (talk) 17:44, 20 April 2009 (UTC)

I've added a short section describing Web APIs (which could use a bit of editing & additional references) because this phrase is now in common parlance, whatever the technical accuracy of describing these services as APIs. Danja (talk) 07:50, 20 July 2009 (UTC)

API did not originally refer to web API's simply because they didn't *exist* at the time. It's absurd to get pedantic and nitpick about the meaning of terms which were never intended to have rigid definitions in the first place. Google themselves refer to their own Maps services as an API -- are we going to say that the guys at Google don't understand computer terminology? Limbo socrates (talk) 22:22, 22 October 2009 (UTC)

I still don't get it

It's late here so maybe it's just me, but I've read the article and I still don't really understand what an API is, or what actual functions it would be used for, and I'm not a stupid person or someone who has no idea about computing at all. People often use the abbreviation in association with Twitter, and I've heard people talk about API limits etc. Someone above has suggested it's too technical and I think maybe they're right... I'll come back tomorrow when I've slept and re-read it and see if I can get my head round it any better. In the meantime if anyone has any expertise in making articles like this more accessible to the average Joe or Joanna, I think it would be useful to look at a bit of a rewrite? 91.107.207.207 (talk) 23:27, 24 April 2009 (UTC)

I skimmed through the main article, and the whole thing looks like it is only intended for people with modern computers, and not computers from the 1980s. Are they talking about tables of symbolic labels that the assembler reads in when processing source code, and churning out object code? Dexter Nextnumber (talk) 04:02, 21 December 2009 (UTC)

Proposed first paragraph

Here's a proposal for a rework of the first paragraph:

In computer science, an application programming interface (API) is a standardized specification for an interface which defines how an application program may request services from the libraries and/or operating systems which provide the services. It may include specifications for routines, data structures, object classes and protocols used to communicate between the requesting software and the software that provides the services. Davemck (talk) 18:48, 20 June 2009 (UTC)

The API - protocol relationship

The article's discussion of how APIs and protocols relate to each other seems confusing and incomplete. Statements like "An API can also be an implementation of a protocol", "API provides a library to be used directly", and "APIs are specific to a given technology" are unclear, most likely inaccurate, possibly even incorrect. In my opinion, it is certain that:

1) There is a relationship between APIs and protocols

2) This relationship is important for describing

2.1) the differences between local (in-process)and remote APIs

2.2) the differences between APIs giving access to the same service or functionality from different programming languages and environments

2.3) the differences between SOAP- and REST-based Web APIs (providing access to identical functionality)

2.4) the precise role of the language bindings

2.5) abstract interfaces which exist only in form of documentation, without explicit or automatically generated language bindings

I don't think this article contains a satisfactory discussion of the above points.

Part of the problem was discussed here before and it has to do with the "wide" versus "narrow" definition of APIs. The "wide" definition includes any interface that provides programmatic access to a service or functionality. The "narrow" definition considers APIs strict in the context of a given programming language, environment or technology. The article starts out by giving the "wide" definition " ...particular set of rules and specifications that software programs can follow to communicate with each other" but the section about protocols treats the subject by taking the "narrow" point of view. This inconsistency is very confusing.

I have a relatively good understanding of the subject, but I'm new to Wikipedia and I don't have references so I'm a bit hesitant to edit the article. Is anybody interested in working with me to clarify the topic?

--Fmihaly (talk) 23:39, 29 June 2011 (UTC)Ferenc

Thank you for your suggestion, I think it will improve the Protocols secion. Feel free to make those changes by yourself. Wikipedia is a wiki, so anyone can edit almost any article by simply following the edit this page link at the top.

The Wikipedia community encourages you to be bold in updating pages. Don't worry too much about making honest mistakes—they're likely to be found and corrected quickly, and other editors can follow the page history to watch changes and add their opinions. If you're not sure how editing works, check out how to edit a page, or use the sandbox to try out your editing skills. New contributors are always welcome. Diego Moya (talk) 09:56, 30 June 2011 (UTC)

Fluent API description removed

I wonder who did remove the part on fluent APIs: I'm editing the "advanced explanation" part of this page since a long time in step-by-step way, and that was a part I was going to develop. So why was it removed? — Preceding unsigned comment added by 62.77.56.17 (talk) 14:39, 25 November 2011 (UTC)

Suggestion of proxy talk

Perhaps mention of the existence of API proxies like Mashery, API Axle (my project), Apigee and 3scale are worth mentioning? — Preceding unsigned comment added by 81.151.214.141 (talk) 21:27, 23 January 2012 (UTC)

Not Copyrightability of APIs

I think there should be a section about the Not Copyrightability of APIs according to the veredict in the "Oracle v. Google" lawsuit ([2]) --Aliuken (talk) 14:40, 1 June 2012 (UTC)