Re: [Squid-cvs] squid3-ipv6 squid3/include config.h,1.7.4.2,1.7.4.3

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Sun, 15 May 2005 21:12:23 +0200 (CEST)

On Sun, 15 May 2005, Rafael Martinez Torres wrote:

> The format of IPv6 URLs are sligthly modified . Since the dotted addresses
> can containg ":" character, it was to mandatory to support brackets to
> split the dotted host address from the port. i.e:
>
> This do not apply to the DNS names...
>
> http://124.35.46.45
> http://124.35.46.45:9090
> http://2004:80:255f:3abb::3
> http://[2004:80:25ab:fe33:5]:9090

Ok. I didn't know there is a defined URL format for IPv6 addresses. Last
time I looked into this it was explicitly excluded, reffering to use of
DNS.

> Hmmm... Yes.. May be URL_NEEDS_BRACKETS ? Any suggestion ?

Fine for me, or perhaps address_needs_brackets?

inline bool address_needs_brackets(const char *host)
{
     return SAFE_INET_ADDR(host,NULL);
}

or perhaps simply

inline bool address_needs_brackets(const char *host)
{
     return strchr(host, ':') != NULL;
}

But in this specific case I think it would make more sense to have
request->host contain the braketed address and make the ipcache (or
SAFE_INET_ADDR) understand that IPv6 addresses may be braketed.

> Yes. For the time beeing, up to the tests I'm not taking care of that.
> My ideas was rather to leave them as macros, yet not defined for
> IPv4... So when the branch becomes steady, it would be easier for the
> group to understand why the compilation fails, and try to merge into head
> in a scalable way...

These kind of macros will be changed into inlines when merged.

> IPv6 branch is not an extension for Squid3, rather a deep change...

I know.

Do you have a time plan when it may be suitable to merge things next?
I hope to soon dive into the 2GB support & cleanup which is another rather
deep going change touching most files little here little there... I hope
to merge your IPv6 bulk changes before this to minimize the risk of
conflicting overlaps. Especially so if you have changes to function
prototypes.

Regards
Henrik
Received on Sun May 15 2005 - 13:12:25 MDT

This archive was generated by hypermail pre-2.1.9 : Tue May 31 2005 - 12:00:03 MDT