Re: I-D Action: draft-carpenter-6man-rfc6874bis-00.txt

[Brian E Carpenter <[e-mail]>]

On 08-Jul-21 21:53, Philip Homburg wrote:
>>> Instead of doing this, maybe we should pick a new character in an update of
>>> RFC 4007 and deprecate '%' as a mistake?
>>
>> Even 9 years ago the WG concluded that '%' was too engrained in code and
>> actual practice to deprecate it. Many libraries that handle IPv6 addresses
>> support '%' and that means that innumerable pieces of software would be
>> impacted, so deprecation seems impractical.
>>
>> If this was a mistake, it was made in 2005 when we approved RFC4007.
> 
> If we started changing 9 years ago, we would be done now. What I see is
> that addresses with zone IDs are used rarely and locally. And lots of software
> doen't even know how to deal with it. 
> 
> If we now introduce, for example, '-' in addition to '%' and require parsers
> to accept both, but canonical output remains at '%' for a while then we
> will have a sensible input format for URLs.
> 
> Keeping '%' as the only option means that URLs will remain a mess.
> 
>>> Aternatively, if we make fe80::ABCD%0 a valid address in the socket API
>>
>> I think you'll find that it already is. I use the netifaces module in Python
>> and it returns things like [{'addr': '00:ff:19:f7:7d:9a'}], 23: [{'addr': 'fe8
>> 0::216a:f0d3:2303:3d1%6', 'netmask': 'ffff:ffff:ffff:ffff::/64', 'broadcast': 
>> 'fe80::ffff:ffff:ffff:ffff%6'}]. It's a bit obscure, though. Also (before
>> Python 3.7) socket.getaddrinfo(socket.gethostname(),0) on Windows would
>> return fe80::216a:f0d3:2303:3d1%6, but somebody changed that behaviour.
> 
> I guess I was too terse.
> 
>>From a socket API point of view writing an IPv6 address without zone ID is
> equivalent to using zone ID 0. So 2001:db8::1 and 2001:db8::1%0 are the
> same. 
> 
> However, in the case of link local, the opengroup specifically disallows
> the use of zone ID 0.

I think that derives from RFC4007:

'  An implementation should also support the concept of a "default" zone
   for each scope.  And, when supported, the index value zero at each
   scope SHOULD be reserved to mean "use the default zone".'

So you can't have a real zone/interface 0. As Michael said, WinSock
got this right (see my message just now). POSIX apparentlt didn't,
but it doesn't matter if you're writing code that is aware of
zone/interface indices.

> 
> If we would introduce a command like
> 'set-default-interface-for-link-local eth0'
> that makes zone ID 0 available, then a URL like http://[fe80::abcd]/
> could work. No change needed to browsers, no complex syntax.

It works today on Windows with Firefox or Chrome.
http://[fe80::2e3a:fdff:fea4:dde7]/ goes straight to my router's
web interface. Fails on Android, which I suppose inherited
Linux's brokenness on this point.

However, this is beside the point for use cases that *need*
to specify the zone/interface index. We'll make doubly sure
that is clear in the next version.

   Brian