query re IPv6 URL format

Brian E Carpenter brian@hursley.ibm.com
Wed, 15 Oct 1997 11:39:27 +0100


I've had that discussion with the URL syntax authors. They are
unanimous that the colons are not parseable in practice.

  Brian

>- Francis Dupont said:
> 
>  In your previous mail you wrote:
> 
>    Brian,
>    
>    >If that was the decision then we need to codify it and tell
>    >the world. I could live with it, if the URL parsers can handle
>    >it. I'll check that with the URL guys.
>    >
>    
>    Maybe I am missing something, but...
>    
>    The IPv4 syntax is:
>    
>    	http://111.222.333.444/~bc
>    
>    It would seem to me that while
>    
>    	http://aaaa:bbbb::cccc:dddd/~bc
>    
>    has multiple ":"'s in it, given left to right parsing, I don't see anything
>    ambiguous.  The first thing after the "//" is either a DNS name, IPv4
>    address, or an IPv6 address.  If the current parsers can distinguish
>    between "hursley.ibm.com" and "111.222.333.444", both of which use dots as
>    separators, then could they also deal with IPv6 addresses with colons?
>    
> => it is not true (just look at the RFC 1738 or for instance lynx
> documentation). According to the last one, URLs can be:
> 
> http://host:port/path?searchpart#fragment
>            ^
> telnet://user:password@host:port
>              ^             ^
> 
> ftp://username:password@host:port/path;type=[D,I, or A]
>               ^             ^
> ...
> And dont' forget we need the same thing for RFC 821/822, X11, ...
> 
>    Why can't the parsers learn how to do this?  They need to be changed to
>    handle IPv6 addresses anyway.
>    
> => it is very easy and as far as I know enough to support DNS names
> (and the terrible RES_USE_INET6 stuff). Personally I think the right
> thing to do is to kill literals (IPv4 literals too)!
> 
> Regards
> 
> Francis.Dupont@inria.fr
> 
> PS: the URL format http://host:port/... is very common and perhaps
> will become more common if ISPs put a bad priority to the 80 TCP port (:-).
>