- -

If the service associated with the processing of a form has side effects (for example, modification of a database or subscription to a service), the method should be POST.

In the HTTP specifications (specifically RFC 2616) the word idempotent is defined as follows:

Methods can also have the property of "idempotence" in that (aside from error or expiration issues) the side-effects of N > 0 identical requests is the same as for a single request.

The word idempotent, as used in this context in the specifications, is (pseudo)mathematical jargon (see definition of "idempotent" in FOLDOC) and should not be taken too seriously or literally here. The phrase "no lasting observable effect on the state of the world" isn''t of course very exact either, and isn''t really the same thing. Idempotent processing, as defined above, does not exclude fundamental changes, only that processing the same data twice has the same effect as processing it once. But here, in fact, idempotent processing means that a form submission causes no changes anywhere except on the user''s screen (or, more generally speaking, in the user agent''s state). Thus, it is basically for retrieving data. If such a form is resubmitted, it might get different data (if the data had been changed meanwhile), but the submission would not cause any update of data or other events. The concept of changes should not be taken too pedantically; for instance, it can hardly be regarded as a change that a form submission is logged into the server''s log file. On the other hand, sending E-mail should normally be regarded as "an effect on the state of the world".

The HTTP specifications aren''t crystal clear on this, and section Safe Methods in the HTTP/1.1 specification describes the principles in yet another way. It opens a different perspective by says that users "cannot be held accountable" for side effects, which presumably means any effect than mere retrieval:

In particular, the convention has been established that the GET and HEAD methods SHOULD NOT have the significance of taking an action other than retrieval. These methods ought to be considered "safe". This allows user agents to represent other methods, such as POST, PUT and DELETE, in a special way, so that the user is made aware of the fact that a possibly unsafe action is being requested.

Naturally, it is not possible to ensure that the server does not generate side-effects as a result of performing a GET request; in fact, some dynamic resources consider that a feature. The important distinction here is that the user did not request the side-effects, so therefore cannot be held accountable for them.

The concept and its background is explained in section Allowing input in Tim Berners-Lee''s Style Guide for online hypertext. It refers, for more information, to User agent watch points, which emphatically says that GET should be used if and only if there are no side effects. But this line of thought, however logical, is not always practical at present, as we shall see.

See also answer to question "What is the difference between GET and POST?" in CGI Programming FAQ by Nick Kew.

Why the distinction matters

In the same context where the pragmatic difference is stated, the HTML 2.0 specification describes the corresponding submission methods technically. However, it does not explain why those methods are recommended for idempotent vs. non-idempotent submissions. Neither does it explain what practical difference there might be between the methods.

Alan Flavell has explained, in an article in a thread titled Limit on URL length in the newsgroup comp.infosystems.www.authoring.html, that "the distinction [between GET and POST] is a real one, and has been observed by some browser makers, which can result in undesirable consequences if the inappropriate one is used". He gives the following reason for conforming to the advice:

When users revisit a page that resulted from a form submission, they might be presented with the page from their history stack (which they had probably intended), or they might be told that the page has now expired. Typical user response to the latter is to hit Reload.

This is harmless if the request is idempotent, which the form author signals to the browser by specifying the GET method.

Browsers typically will (indeed "should") caution their users if they are about to resubmit a POST request, in the belief that this is going to cause a further "permanent change in the state of the universe", e.g. ordering another Mercedes-Benz against their credit card or whatever. If users get so accustomed to this happening when they try to reload a harmless idempotent request, then sooner or later it''s going to bite them when they casually [OK] the request and do, indeed, order a second pizza, or invalidate their previous competition entry by apparently trying to enter twice, or whatever.

Thus, some browsers can act more cleverly if the author uses "GET" or "POST" consistently, i.e. using "GET" for pure queries and "POST" for other form submissions. It needs to be noted, though, that using "GET" gives no protection against causing changes. A script which processes a form submission sent with the "GET" could cause a pizza ordering. It''s just that authors are expected to take care that such things don''t happen.

Moreover, the use of "POST" cannot guarantee that the user does not inadvertantly submit the same form data twice; the browser might not give a warning, or the user might fail to understand the warning. Users are known to become impatient when it seems that "nothing happens" when they click on a button, so they might click on it again and again. Thus, robust processing of forms should take precautions against unintended duplicate actions. (As a simple example, a submission might be processed first by a script which sends back a page containing a confirmation request, echoing back the data submitted and asking the user to verify it and then submit the confirmation.)

A "GET" request is often cacheable, whereas a "POST" request can hardly be. For query systems this may have a considerable efficiency impact, especially if the query strings are simple, since caches might serve the most frequent queries. For information about caches, see Caching Tutorial for Web Authors and Webmasters, especially section Writing Cache-Aware Scripts.

Differences in form submission

For both METHOD="GET" and METHOD="POST", the processing of a user''s submit request (such as clicking on a submit button) in a browser begins with a construction of the form data set, which is then encoded in a manner which depends on the ENCTYPE attribute. That attribute has two possible values mentioned in the specifications, but multipart/form-data is for "POST" submissions only, whereas application/x-www-form-urlencoded (the default) can be used both for "POST" and for "GET".

Then the form data set is transmitted as follows (quotation from the HTML 4.0 specification):

• If the method is "get" - -, the user agent takes the value of action, appends a ? to it, then appends the form data set, encoded using the application/x-www-form-urlencoded content type. The user agent then traverses the link to this URI. In this scenario, form data are restricted to ASCII codes.
• If the method is "post" --, the user agent conducts an HTTP post transaction using the value of the action attribute and a message created according to the content type specified by the enctype attribute.

Thus, for METHOD="GET" the form data is encoded into a URL (or, speaking more generally, into a URI). This means that an equivalent to a form submission can be achieved by following a normal link referring to a suitable URL; see the document Choices in HTML forms for details and examples. On a typical browser, the user sees the URL of a document somewhere (e.g. on Location line), and if he is viewing the results of a query sent using METHOD="GET", he will see what the actual query was (i.e. the part of the URL that follows the ? sign). The user could then bookmark it or cut&paste it for later use (e.g. to be E-mailed or put into one''s own HTML document after some editing).

Although the HTML specifications don''t say it very explicitly, the fundamental difference between the methods is really that they correspond to different HTTP requests, as defined in the HTTP specifications. See especially Method Definitions in RFC 2616. For form submission with METHOD="GET", the browser constructs a URL as described above, then processes it as if following a link (or as if the user had typed the URL directly). The browser divides the URL into parts and recognizes a host, then sends to that host a GET request with the rest of the URL as argument. The server takes it from there. Submission of a form with METHOD="POST" causes a POST request to be sent.

Differences in server-side processing

In principle, processing of a submitted form data depends on whether it is sent with METHOD="GET" or METHOD="POST". Since the data is encoded in different ways, different decoding mechanisms are needed. Thus, generally speaking, changing the METHOD may necessitate a change in the script which processes the submission. For example, when using the CGI interface, the script receives the data in an environment variable when METHOD="GET" is used but in the standard input stream (stdin) when METHOD="POST" is used.

It is, however, possible to construct libraries of subroutines (e.g. Perl functions) which allow one to write scripts in a manner which works both for METHOD="GET" and METHOD="POST". This would be based on distinguishing between the cases within the subroutine code and returning the data to the caller in a uniform manner.

Possible reasons to use "POST" for idempotent queries

For reasons explained above, one should normally use METHOD="POST" if and only if the form submission may cause changes. There are some exceptional situations where one may consider using METHOD="POST" even for pure queries, too:

• If the form data would contain non-ASCII characters, then METHOD="GET" is inapplicable in principle, although it may work in practice (mainly for ISO Latin 1 characters). Thus, for a query where the keywords might contain e.g. accented letters, you have to select among two evils: using METHOD="GET" against the rules which restrict the character reportoire to ASCII within it, or using METHOD="POST" against the rules which says that it should not be used when the processing is idempotent. The latter alternative is probably less dangerous.
• If the form data set is large - say, hundreds of characters - then METHOD="GET" may cause practical problems with implementations which cannot handle that long URLs. Such usage is mentioned in the the HTML 2.0 specification in an informative note as follows: Note - The URL encoding may result in very long URIs, which cause some historical HTTP server implementations to exhibit defective behavior. As a result, some HTML forms are written using METHOD=POST even though the form submission has no side-effects. The limitations are not only historical. There is an official statement by Microsoft, originally published 2000-02-23: INFO: Maximum URL Length Is 2,083 Characters in Internet Explorer (Q208427).
• You might wish to avoid METHOD="GET" in order to make it less visible to users how the form works, especially in order to make "hidden" fields (INPUT TYPE="HIDDEN") more hidden. Using POST implies that users won''t see the form data in the URL shown by the user; but note that this is not a very effective method of hiding, since the user can of course view the source code of your FORM element!