Valid characters for URI schemes?

Question

I was thinking about Registering an Application to a URL Protocol and I'd like to know, what characters are allowed in a scheme?

Some examples:

• h323 (has numbers)
  • h323:[<user>@]<host>[:<port>][;<parameters>]
• z39.50r (has a . as well)
  • z39.50r://<host>[:<port>]/<database>?<docid>[;esn=<elementset>][;rs=<recordsyntax>]
• paparazzi:http (has a :)
  • paparazzi:http:[//<host>[:[<port>][<transport>]]/

So, what characters can I fancy using?
Can we have...

• @:TwitterUser
• #:HashTag
• $:CapitalStock
• ?:ID-10T

...etc., as desired, or characters in the scheme are restricted by standard?

Abyssal

Answer 1

According to RFC 2396, Appendix A:

  scheme        = alpha *( alpha | digit | "+" | "-" | "." )

Meaning:

The scheme should start with a letter (upper or lower case), and can contains letters (still upper and lower case), number, "+", "-" and ".".

---

Note: in the case of

paparazzi:http:[//<host>[:[<port>][<transport>]]/

the scheme is only the "paparazzi" part.

Maharani

I see. But there are RFCs that use numbers... Why? – Abyssal

Numbers are allowed in the URI scheme, but not as first character. 'a234' is valid, while '4bcd' isn't. – Maharani

Do you think the fact that it will be used only as an URL protocol on Windows has any impact on the usability of other characters? – Abyssal

+1; What Vivien said re: the "paparazzi:" scheme. The http://... is passed on to the WebKit stuff. (NB: I'm the author of the app, and am also crazy and have a BNF-style document on the URL format on the site.) – Incult

paparazzi is akin to mailto: it has no hierarchy hence no // – Skippie

Chiming in 11 years later on Camilo's question: Windows does not enforce the starts-with-alpha limitation, but because all popular browsers do, you should follow it anyway. – Guiscard

Answer 2

The scheme according to RFC 3986 is defined as:

scheme      = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." )

So the scheme must begin with an alphabetic character (A–Z, a–z) and may be followed by any number of alphanumeric characters, +, -, or ..

Ludwigshafen

Do you think using it as a Windows-only URL protocol has any impact on the characters used? If that changes anything I'd do some tests... – Abyssal

Answer 3

Quoth RFC 2396:

Scheme names consist of a sequence of characters beginning with a lower case letter and followed by any combination of lower case letters, digits, plus ("+"), period ("."), or hyphen ("-").

Lightly