The generic URI syntax actually specifies that any path segment within a URI can be divided up into parts by semicolons (or commas), and suggests using key=value as parameters within that segment.

http://tools.ietf.org/html/rfc3986#section-3.3

   For
   example, the semicolon (";") and equals ("=") reserved characters are
   often used to delimit parameters and parameter values applicable to
   that segment.  The comma (",") reserved character is often used for
   similar purposes.  For example, one URI producer might use a segment
   such as "name;v=1.1" to indicate a reference to version 1.1 of
   "name", whereas another might use a segment such as "name,1.1" to
   indicate the same

There’s no thought given as far as I can see to the relative URL concepts this matrix proposal covers, but the idea of adding key=value data in this format is supported by the RFC, and indeed in a way that supports further adding sub resources below the current one.

Which, by the way, this matrix relative URL proposal seems to ignore - how is ./subpage resolved relative to /map;x=250;y=582 ?

Many comments so far seem to imply (perhaps not intentionally) that this syntax was intended only on the "right side" of the URI, but you can think of it more as adding qualifier-like information to (potentially) each path segment. So in /com/foo/bar you could further qualify foo (perhaps as /com/foo;color=yellow/bar). This can be combined with a (definitionally trailing) query string as well. Someone noted that Spring implements some support for this; the Eclipse Foundation's Jersey (Java REST toolkit) does as well if I remember right.

In every monolithic Java Jersey server I worked in there was always at least 1 class using @MatrixParam instead of the query param equivalent. No one could ever answer me why or who because it was always written in a project before it migrated to git.

Semicolon is also nice on the eyes compared to '&' we're used to today.

Why in the world this didn't gain traction?

Alan Kay kind of said something along those lines, in that he envisioned the world wide web to be more integrative.

If we include this with Matrix URIs, we could have a scheme format that could work with remote data/objects in a convenient and logical way. RESTful APIs do not seem to be quite the same though. Like some universal API ... not sure how it would look.

If my query would be "weather?" then any API should understand this to mean I'd like to return the local weather. But as an API. Naturally we have AI now so any such queries become trivial at all times, but this is still not quite what was meant. We kind of limited things here because "all external input may be malicious". This then meant no further development in this area, because "too dangerous". Every time I try to use JavaScript to read a local file, I am reminded of that as limitation. They did not think that there are valid, non-malicious use cases, so I am using ruby and python rather than JavaScript (I could use node, but it kind of feels ... so dirty when using it).

I quite like the semi-colons. I wonder how we ended up with the slightly unwieldy ?foo=bar&some=thing

Think “how do I implement a search engine whose interface is visible in the browser address bar?” This was a reasonable question on the early web. And that’s why they are called query params.

The question mark indicates you’re making a query. And the ampersand is a boolean operator. You could imagine query params implementing | in addition to &.

I think to non-techies the semi-colons are still more intuitive - I often see non-technical people mess up the ? and &, either by using just one of them or by getting them the wrong way around.

I chuckled once when driving having a radio adverts announce "enter in your web browser something.com slash question-mark prize ampersand win to win!"

I'm certain I've seen websites which used semi-colons rather than question marks and ampersands for query parameters. The Jehovah's Witnesses website used to, many years ago. I believe that one of the reasons behind this design was that it meant that there were no special characters which needed to be escaped in HTML. This design works fine for links, but not for forms, which will generate normal URLs containing question marks and ampersands. I suppose you can use POST-Redirect-GET forms, and generate your own URLs using semi-colons.

Seems Like the recommendation changed from supporting ; and & to only using &.

Thinking about it, it is a little surprising as, if I remember correctly, in HTML source you should encode & as & right?

& as an escape character only applies to text nodes. Of course, if you want to display the URL on a page, you have to escape it, but not in the href.

You're supposed to escape & anywhere in HTML, not just in text nodes. If you don't, it'll probably work, but browser first tries to interpret it as a start of an entity anyway. Even if it is inside a href etc.

If you're relying on HTML forms being filled in to generate queries, the browser will always generate standard query parameters with a question mark and ampersands. But if you're making your own links, you can generate them however you like. Semi-colons should work fine, though you may need to write your one URL parsing functions.

Iirc in java I managed to parse those just fine. Ages ago though.

I also recall actually seeing some web pages somewhere using semicolon, possibly IBM stuff.

I think some browsers display the query params muted compared to the part before (or maybe I’m confusing it with the behavior that the domain is highlighted to prevent phishing?). But technically, you can just parse the request url however you want. I also think it shouldn’t matter with regards to any caching layers inbetween.