Semicolon as URL query separator

Question

Although it is strongly recommended (W3C source, via Wikipedia) for web servers to support semicolon as a separator of URL query items (in addition to ampersand), it does not seem to be generally followed.

For example, compare

        http://www.google.com/search?q=nemo&oe=utf-8

        http://www.google.com/search?q=nemo;oe=utf-8

results. (In the latter case, semicolon is, or was at the time of writing this text, treated as ordinary string character, as if the url was: http://www.google.com/search?q=nemo%3Boe=utf-8)

Although the first URL parsing library i tried, behaves well:

>>> from urlparse import urlparse, query_qs
>>> url = 'http://www.google.com/search?q=nemo;oe=utf-8'
>>> parse_qs(urlparse(url).query)
{'q': ['nemo'], 'oe': ['utf-8']}

What is the current status of accepting semicolon as a separator, and what are potential issues or some interesting notes? (from both server and client point of view)

Answer 1

The W3C Recommendation from 1999 is obsolete. The current status, according to the 2014 W3C Recommendation, is that semicolon is now illegal as a parameter separator:

To decode application/x-www-form-urlencoded payloads, the following algorithm should be used. [...] The output of this algorithm is a sorted list of name-value pairs. [...]

1. Let strings be the result of strictly splitting the string payload on U+0026 AMPERSAND characters (&).

In other words, ?foo=bar;baz means the parameter foo will have the value bar;baz; whereas ?foo=bar;baz=sna should result in foo being bar;baz=sna (although technically illegal since the second = should be escaped to %3D).

Answer 2

As long as your HTTP server, and your server-side application, accept semicolons as separators, you should be good to go. I cannot see any drawbacks. As you said, the W3C spec is on your side:

We recommend that HTTP server implementors, and in particular, CGI implementors support the use of ";" in place of "&" to save authors the trouble of escaping "&" characters in this manner.

Answer 3

I agree with Bob Aman. The W3C spec is designed to make it easier to use anchor hyperlinks with URLs that look like form GET requests (e.g., http://www.host.com/?x=1&y=2). In this context, the ampersand conflicts with the system for character entity references, which all start with an ampersand (e.g., "). So W3C recommends that web servers allow a semicolon to be used as a field separator instead of an ampersand, to make it easier to write these URLs. But this solution requires that writers remember that the ampersand must be replaced by something, and that a ; is an equally valid field delimiter, even though web browsers universally use ampersands in the URL when submitting forms. That is arguably more difficult that remembering to replace the ampersand with an & in these links, just as would be done elsewhere in the document.

To make matters worse, until all web servers allow semicolons as field delimiters, URL writers can only use this shortcut for some hosts, and must use & for others. They will also have to change their code later if a given host stops allowing semicolon delimiters. This is certainly harder than simply using &, which will work for every server forever. This in turn removes any incentive for web servers to allow semicolons as field separators. Why bother, when everyone is already changing the ampersand to & instead of ;?

Answer 4

In short, HTML is a big mess (due to its leniency), and using semicolons help to simplify this a LOT. I estimate that when i factor in the complications that i've found, using ampersands as a separator makes the whole process about three times as complicated as using semicolons for separators instead!

I'm a .NET programmer and to my knowledge, .NET does not inherently allow ';' separators, so i wrote my own parsing and handling methods because i saw a tremendous value in using semicolons rather than the already problematic system of using ampersands as separators. Unfortunately, very respectable people (like @Bob Aman in another answer) do not see the value in why semicolon usage is far superior and so much simpler than using ampersands. So i now share a few points to perhaps persuade other respectable developers who don't recognize the value yet of using semicolons instead:

Using a querystring like '?a=1&b=2' in an HTML page is improper (without HTML encoding it first), but most of the time it works. This however is only due to most browsers being tolerant, and that tolerance can lead to hard-to-find bugs when, for instance, the value of the key value pair gets posted in an HTML page URL without proper encoding (directly as '?a=1&b=2' in the HTML source). A QueryString like '?who=me+&+you' is problematic too.

We people can have biases and can disagree about our biases all day long, so recognizing our biases is very important. For instance, i agree that i just think separating with ';' looks 'cleaner'. I agree that my 'cleaner' opinion is purely a bias. And another developer can have an equally opposite and equally valid bias. So my bias on this one point is not any more correct than the opposite bias.

But given the unbiased support of the semicolon making everyone's life easier in the long run, cannot be correctly disputed when the whole picture is taken into account. In short, using semicolons does make life simpler for everyone, with one exception: a small hurdle of getting used to something new. That's all. It's always more difficult to make anything change. But the difficulty of making the change pales in comparison to the continued difficulty of continuing to use &.

Using ; as a QueryString separator makes it MUCH simpler. Ampersand separators are more than twice as difficult to code properly than if semicolons were used. (I think) most implementations are not coded properly, so most implementations aren't twice as complicated. But then tracking down and fixing the bugs leads to lost productivity. Here, i point out 2 separate encoding steps needed to properly encode a QueryString when & is the separator:

• Step 1: URL encode both the keys and values of the querystring.
• Step 2: Concatenate the keys and values like 'a=1&b=2' after they are URL encoded from step 1.
• Step 3: Then HTML encode the whole QueryString in the HTML source of the page.

So special encoding must be done twice for proper (bug free) URL encoding, and not just that, but the encodings are two distinct, different encoding types. The first is a URL encoding and the second is an HTML encoding (for HTML source code). If any of these is incorrect, then i can find you a bug. But step 3 is different for XML. For XML, then XML character entity encoding is needed instead (which is almost identical). My point is that the last encoding is dependent upon the context of the URL, whether that be in an HTML web page, or in XML documentation.

Now with the much simpler semicolon separators, the process is as one wud expect:

• 1: URL encode the keys and values,
• 2: concatenate the values together. (With no encoding for step 3.)

I think most web developers skip step 3 because browsers are so lenient. But this leads to bugs and more complications when hunting down those bugs or users not being able to do things if those bugs were not present, or writing bug reports, etc.

Another complication in real use is when writing XML documentation markup in my source code in both C# and VB.NET. Since & must be encoded, it's a real drag, literally, on my productivity. That extra step 3 makes it harder to read the source code too. So this harder-to-read deficit applies not only to HTML and XML, but also to other applications like C# and VB.NET code because their documentation uses XML documentation. So the step #3 encoding complication proliferates to other applications too.

So in summary, using the ; as a separator is simple because the (correct) process when using the semicolon is how one wud normally expect the process to be: only one step of encoding needs to take place.

Perhaps this wasn't too confusing. But all the confusion or difficulty is due to using a separation character that shud be HTML encoded. Thus '&' is the culprit. And semicolon relieves all that complication.

(I will point out that my 3 step vs 2 step process above is usually how many steps it would take for most applications. However, for completely robust code, all 3 steps are needed no matter which separator is used. But in my experience, most implementations are sloppy and not robust. So using semicolon as the querystring separator would make life easier for more people with less website and interop bugs, if everyone adopted the semicolon as the default instead of the ampersand.)