W3C home > Mailing lists > Public > www-talk@w3.org > January to February 2003

Re: http://///////////////

From: Ian Clelland <ian@veryfresh.com>
Date: Wed, 5 Feb 2003 08:54:32 +0000
To: Diwakar Shetty <diwakar.shetty@oracle.com>
Cc: www-talk@w3.org
Message-ID: <20030205085432.G19695@fullfactor.com>

On Wed, Feb 05, 2003 at 12:04:08PM +0530, Diwakar Shetty wrote:
> Is the following valid HTTP ??
> 
> telnet web_server_host http_port
> GET http://///////////////    HTTP/1.0

I think that this would be syntactically valid HTTP, at least 
according to the definition of 'absoluteURI' in RFC1945 (S3.2.1), 
*except* that the scheme is specified as 'http', and it violates 
S3.2.2, since a hostname is required (and it must be at least one 
character long).

However, I believe that all of these close alternatives are valid:

GET /////////////// HTTP/1.0

(since /////////////// is a valid relativeURI)


GET http://a/////////////// HTTP/1.0

(since 'a' is a valid hostname, http://a/////////////// becomes a 
valid http URI)


GET xyz://///////////// HTTP/1.0

(since the 'xyz' schema is not defined anywhere, this is a valid 
absoluteURI, although no HTTP server would know what to do with it)


Can I ask what inspired the question in the first place? Is this a real-world discovery, or just an excercise in pathological URIs?


Regards,

Ian Clelland
<ian@veryfresh.com>
Received on Wednesday, 5 February 2003 03:54:34 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 27 October 2010 18:14:27 GMT