XML and JSON are the two most common formats for data interchange in the Web today. XML was created by the W3C in 1996, and JSON was publicly specified by Douglas Crockford in 2002. Although their purposes are not identical, they are frequently used to accomplish the same task, which is data interchange. Both have well-documented open standards on the Web ([RFC 7159][1], [RFC 4825][2]), and both are human and machine-readable. Neither one is absolutely superior to the other, as each is better suited for different use cases.
There are several advantages that XML has over JSON. One of the biggest differences between the two is that in XML you can put metadata into the tags in the form of attributes. With JSON, the programmer could accomplish the same goal that metadata achieves by making the entity an object and adding the attributes as members of the object. However, the way XML does it may often be preferable, given the slightly misleading nature of turning something into an object that is not one in the client program. For example, if your C++ program sends an int via JSON and needs metadata to be sent along with it, you would have to make it an object, with one name/value pair for the actual value of the int, and more name/value pairs for each attribute. The program that receives the JSON would read it as an object, when in fact it is not one. While this is a viable solution, it defies one of JSON’s key advantages: “JSON's structures look like conventional programming language structures. No restructuring is necessary.”[2]
XML 与 JSON 相比有很多优点。二者间最大的不同在于使用 XML 时可以通过在标签中添加属性这一简单的方法来存储<ruby>元数据<rt>metadata</rt></ruby>。而使用 JSON 时需要创建一个对象，把元数据当作对象的成员来存储。虽然二者都能达到存储元数据的目的，但在这一情况下通常选择 XML，因为 JSON 的表达形式会让客户端程序误以为要将数据转换成一个对象。举个例子，如果你的 C++ 程序需要使用 JSON 格式发送一个附带元数据的整型数据，需要创建一个对象，用对象中的一个<ruby>名称/值对<rt>name/value pair</rt></ruby>来记录整型数据的值，再为每一个附带的属性添加一个名称/值对。接收到这个 JSON 的程序在读取后很可能把它当成一个对象，可事实并不是这样。虽然这是使用 JSON 传递元数据的一种变通方法，但他违背了 JSON 的核心优势：“<ruby>JSON 的结构与常规的程序语言中的结构相对应，而无需修改。<rt>JSON's structures look like conventional programming language structures. No restructuring is necessary.</rt></ruby>”[2]
Although I will argue later that this can also be a drawback of XML, its mechanism to resolve name conflicts, prefixes, gives it power that JSON does not have. With prefixes, the programmer has the ability to name two different kinds of entities the same thing.[1] This would be advantageous in situations where the different entities should have the same name in the client program, perhaps if they are used in entirely different scopes.
Another advantage of XML is that most browsers render it in a highly readable and organized way. The tree structure of XML lends itself well to this formatting, and allows for browsers to let users to naturally collapse individual tree elements. This feature would be particularly useful in debugging.
One of the most significant advantages that XML has over JSON is its ability to communicate mixed content, i.e. strings that contain structured markup. In order to handle this with XML, the programmer need only put the marked-up text within a child tag of the parent in which it belongs. Similar to the metadata situation, since JSON only contains data, there is no such simple way to indicate markup. It would again require storing metadata as data, which could be considered an abuse of the format.
XML 对比 JSON 有一个很重要的能力就是它可以<ruby>混合多种内容<rt>mixed content</rt></ruby>。例如在 XML 中处理包含结构化标记的字符串时，程序员们只要把带有标记的文本放在一个标签内就可以了。而且和元数据的处理类似，由于 JSON 只包含数据，没有用于指明标签的简单方式。虽然这里还可以使用处理元数据的解决方法，但这总有点误用格式之嫌。
JSON has several advantages as well. One of the most obvious of these is that JSON is significantly less verbose than XML, because XML necessitates opening and closing tags (or in some cases less verbose self-closing tags), and JSON uses name/value pairs, concisely delineated by “{“ and “}” for objects, “[“ and “]” for arrays, “,” to separate pairs, and “:” to separate name from value. Even when zipped (using gzip), JSON is still smaller and it takes less time to zip it.[6] As determined by Sumaray and Makki as well as Nurseitov, Paulson, Reynolds, and Izurieta in their experimental findings, JSON outperforms XML in a number of ways. First, naturally following from its conciseness, JSON files that contain the same information as their XML counterparts are almost always significantly smaller, which leads to faster transmission and processing. Second, difference in size aside, both groups found that JSON was serialized and deserialized drastically faster than XML.[3][4] Third, the latter study determined that JSON processing outdoes XML in CPU resource utilization. They found that JSON used less total resources, more user CPU, and less system CPU. The experiment used RedHat machines, and RedHat claims that higher user CPU usage is preferable.[3] Unsurprisingly, the Sumaray and Makki study determined that JSON performance is superior to XML on mobile devices too.[4] This makes sense, given that JSON uses less resources, and mobile devices are less powerful than desktop machines.
Yet another advantage that JSON has over XML is that its representation of objects and arrays allows for direct mapping onto the corresponding data structures in the host language, such as objects, records, structs, dictionaries, hash tables, keyed lists, and associative arrays for objects, and arrays, vectors, lists, and sequences for arrays.[2] Although it is perfectly possible to represent these structures in XML, it is only as a function of the parsing, and it takes more code to serialize and deserialize properly. It also would not always be obvious to the reader of arbitrary XML what tags represent an object and what tags represent an array, especially because nested tags can just as easily be structured markup instead. The curly braces and brackets of JSON definitively show the structure of the data. However, this advantage does come with the caveat explained above, that the JSON can inaccurately represent the data if the need arises to send metadata.
JSON 的另一个优点在于其对对象和数组的描述允许<ruby>宿主语言<rt>host language</rt></ruby>直接将它映射到对应数据结构上，例如<ruby>对象<rt>object</rt></ruby>、<ruby>记录<rt>record</rt></ruby>、<ruby>结构体<rt>struct</rt></ruby>、<ruby>字典<rt>dictionary</rt></ruby>、<ruby>哈希表<rt>hash table</rt></ruby>、<ruby>键值列表<rt>keyed list</rt></ruby>还有对象组成的数组，以及<ruby>数组<rt>array</rt></ruby>、<ruby>向量<rt>vector</rt></ruby>、<ruby>列表<rt>list</rt></ruby>等等。[2] 虽然 XML 里也能表达这些数据结构，也只需调用一个函数就能完成解析，但需要更多的代码才能正确的完成 XML 的序列化和反序列化处理。而且 XML 对于人类来说不如 JSON 那么直观，因为 XML 标准缺乏对象、数组的标签的明确定义，尤其是嵌套的标签可以简单的使用结构化的标记替代时。JSON 中的花括号和中括号则明确表示了数据的结构，当然这一优势也符合前文中的警告，在包含元数据时 JSON 的表示不如 XML 精确。
Although XML supports namespaces and prefixes, JSON’s handling of name collisions is less verbose than prefixes, and arguably feels more natural with the program using it; in JSON, each object is its own namespace, so names may be repeated as long as they are in different scopes. This may be preferable, as in most programming languages members of different objects can have the same name, because they are distinguished by the names of the objects to which they belong.
Perhaps the most significant advantage that JSON has over XML is that JSON is a subset of JavaScript, so code to parse and package it fits very naturally into JavaScript code. This seems highly beneficial for JavaScript programs, but does not directly benefit any programs that use languages other than JavaScript. However, this drawback has been largely overcome, as currently the JSON website lists over 175 tools for 64 different programming languages that exist to integrate JSON processing. While I cannot speak to the quality of most of these tools, it is clear that the developer community has embraced JSON and has made it simple to use in many different platforms.
Simply put, XML’s purpose is document markup. This is decidedly not a purpose of JSON, so XML should be used whenever this is what needs to be done. It accomplishes this purpose by giving semantic meaning to text through its tree-like structure and ability to represent mixed content. Data structures can be represented in XML, but that is not its purpose.
JSON’s purpose is structured data interchange. It serves this purpose by directly representing objects, arrays, numbers, strings, and booleans. Its purpose is distinctly not document markup. As described above, JSON does not have a natural way to represent mixed content.
Of the top 10 most popular APIs according to Programmable Web[9], along with a couple more popular ones, only one supports XML and not JSON. Several support both, and several support only JSON. Among developer APIs for modern and popular websites, JSON clearly seems to be preferred. This also indicates that more app developers that use these APIs prefer JSON. This is likely a result of its reputation as the faster and leaner of the two. Furthermore, most of these APIs communicate data rather than documents, so JSON would be more appropriate. For example, Facebook is mainly concerned with communicating data about users and posts, Google Maps deals in coordinates and information about entities on their maps, and AccuWeather just sends weather data. Overall, it is impossible to say whether JSON or XML is currently used more in APIs, but the trend is certainly swinging towards JSON.[10][11]
根据<ruby>可编程网络<rt>Programmable Web</rt></ruby>[9]的数据，最流行的 10 个 API 中只有一个是仅提供 XML 且不支持 JSON 的。其他的要么同时支持 XML 和 JSON，要么只支持 JSON。这表明了大多数应用开发者都更倾向于使用支持 JSON 的 API，原因大概是因为 JSON 的处理更快的良好口碑和与 XML 相比相对轻量。此外，大多数 API 只是传递数据而非文档，所以 JSON 更加合适。例如 Facebook 的重点在于用户的交流与帖子，谷歌地图则主要处理坐标和地图信息，AccuWeather 就只传递天气数据。总之，虽然不能说天气 API 在使用时究竟是 JSON 用的多还是 XML 用的多，但是趋势明确偏向了 JSON。[10][11]
It makes sense for software that is mainly concerned with document creation, manipulation, and storage to use XML rather than JSON. Also, all three of these programs support mixed content, which JSON does not do well. For example, if a user is typing up an essay in Microsoft Word, they may put different font, size, color, positioning, and styling on different blocks of text. XML naturally represents these properties with nested tags and attributes.
因为这些软件需要考虑引用、格式、存储等等，所以比起 JSON，XML 优势更大。另外，这三款程序都支持混合内容，而 JSON 在这一点上做得并不如 XML 好。举例说明，当用户使用 Microsoft Word 编辑一篇论文时，用户需要使用不同的文字字形、文字大小、文字颜色、页边距、段落格式等，而 XML 结构化的组织形式与标签属性生来就是为了表达这些信息的。
For a long time, SQL and the relational database model dominated the market. Corporate giants like Oracle and Microsoft have always marketed such databases. However, in the last decade, there has been a major rise in popularity of NoSQL databases. As this has coincided with the rise of JSON, most NoSQL databases support JSON, and some, such as MongoDB, CouchDB, and Riak use JSON to store their data. These databases have two important qualities that make them better suited for modern websites: they are generally more scalable than relational SQL databases, and they are designed to the core to run in the Web.[10] Since JSON is more lightweight and a subset of JavaScript, it suits NoSQL databases well, and helps facilitate these two qualities. In addition, many older databases have added support for JSON, such as Oracle Database and PostgresSQL. Conversion between XML and JSON is a hassle, so naturally, as more developers use JSON for their apps, more database companies have incentive to support it.[7]
One of the most heavily anticipated changes in the Internet is the “Internet of Things”, i.e. the addition to the Internet of non-computer devices such as watches, thermostats, televisions, refrigerators, etc. This movement is well underway, and is expected to explode in the near future, as predictions for the number of devices in the Internet of Things in 2020 range from about 26 billion to 200 billion.[12][13][13] Almost all of these devices are smaller and less powerful than laptop and desktop computers. Many of them only run embedded systems. Thus, when they have the need to exchange data with other entities in the Web, the lighter and faster JSON would naturally be preferable to XML.[10] Also, with the recent rapid rise of JSON use in the Web relative to XML, new devices may benefit more from speaking JSON. This highlights an example of Metcalf’s Law; whether XML or JSON or something entirely new becomes the most popular format in the Web, newly added devices and all existing devices will benefit much more if the newly added devices speak the most popular language.
对互联网的变革中，最让人期待的便是<ruby>物联网<rt>Internet of Things</rt></ruby>。这会给互联网带来大量的非计算机设备，例如手表、温度计、电视、冰箱等等。这一势头的发展良好，预期将在不久的将来迎来爆发式的增长。据估计，到 2020 年时会有 260 亿 到 2000 亿的物联网设备被接入互联网。[12][13] 几乎所有的物联网设备都是小型设备，此外比笔记本或者台式电脑的性能要弱很多。大多数都是嵌入式系统。因此，当他们需要与 web 上的系统交换数据时，更轻量，更快速的 JSON 自然比 XML 更受青睐。[10] 受益于 JSON 在 web 上的快速普及，与 XML 相比，这些新的物联网设备更有可能从使用 JSON 中受益。这是一个典型的梅特卡夫定律的实例，无论是 XML 还是 JSON，抑或是什么其他全新的格式，现存的设备和新的设备都会从支持最广泛使用的格式中受益。
With the creation and recent rapid increase in popularity of Node.js, a server-side JavaScript framework, along with NoSQL databases like MongoDB, full-stack JavaScript development has become a reality. This bodes well for the future of JSON, as with these new apps, JSON is spoken at every level of the stack, which generally makes the apps very fast and lightweight. This is a desirable trait for any app, so this trend towards full-stack JavaScript is not likely to die out anytime soon.[10]
Another existing trend in the world of app development is toward REST and away from SOAP.[11][15][16] Both XML and JSON can be used with REST, but SOAP exclusively uses XML.
The given trends indicate that JSON will continue to dominate the Web, and XML use will continue to decrease. This should not be overblown, however, because XML is still very heavily used in the Web, and it is the only option for apps that use SOAP. Given the widespread migration from SOAP to REST, the rise of NoSQL databases and full-stack JavaScript, and the far superior performance of JSON, I believe that JSON will soon be much more widely used than XML in the Web. There seem to be very few applications where XML is the better choice.
3. [Comparison of JSON and XML Data Interchange Formats: A Case Study](http://www.cs.montana.edu/izurieta/pubs/caine2009.pdf)
4. [A comparison of data serialization formats for optimal efficiency on a mobile platform](http://dl.acm.org/citation.cfm?id=2184810)
5. [JSON vs. XML: The Debate](http://ajaxian.com/archives/json-vs-xml-the-debate)
6. [JSON vs. XML: Some hard numbers about verbosity](http://www.codeproject.com/Articles/604720/JSON-vs-XML-Some-hard-numbers-about-verbosity)
7. [How JSON sparked NoSQL -- and will return to the RDBMS fold](http://www.infoworld.com/article/2608293/nosql/how-json-sparked-nosql----and-will-return-to-the-rdbms-fold.html)
8. [Did You Say "JSON Support" in Oracle 12.1.0.2?](https://blogs.oracle.com/OTN-DBA-DEV-Watercooler/entry/did_you_say_json_support)
9. [Most Popular APIs: At Least One Will Surprise You](http://www.programmableweb.com/news/most-popular-apis-least-one-will-surprise-you/2014/01/23)
10. [Why JSON will continue to push XML out of the picture](https://www.centurylinkcloud.com/blog/post/why-json-will-continue-to-push-xml-out-of-the-picture/)
11. [Thousands of APIs Paint a Bright Future for the Web](http://www.webmonkey.com/2011/03/thousand-of-apis-paint-a-bright-future-for-the-web/)
12. [A Simple Explanation Of 'The Internet Of Things’](http://www.forbes.com/sites/jacobmorgan/2014/05/13/simple-explanation-internet-things-that-anyone-can-understand/)
13. [Proofpoint Uncovers Internet of Things (IoT) Cyberattack](http://www.proofpoint.com/about-us/press-releases/01162014.php)
14. [The Internet of Things: New Threats Emerge in a Connected World](http://www.symantec.com/connect/blogs/internet-things-new-threats-emerge-connected-world)
15. [3,000 Web APIs: Trends From A Quickly Growing Directory](http://www.programmableweb.com/news/3000-web-apis-trends-quickly-growing-directory/2011/03/08)
16. [How REST replaced SOAP on the Web: What it means to you](http://www.infoq.com/articles/rest-soap)
