Zero prefix considered invalid for IPv6

[ioprev]

When trying to create an IPAddress object for the IPv6 address ::/0 the following error occur:
ArgumentError: Prefix must be in range 1..128, got: 0

This is a result of a check done at line 217 of ipaddress/prefix.rb file.

The address ::/0 is a special case of an IPv6 address and should be considered a valid IPv6 address representation. See: http://en.wikipedia.org/wiki/IPv6_address#Default_route

[crosson]

That is a valid route entry. Is it a valid ip address? In the case of ipv4 it isn't.

[crosson]

Huh. On the flip side Prefix32 does not appear to be consistent Prefix128 since it does allow the 0.

[bluemonk]

Good point.
I think I will enable the /0 on Prefix128 since I could not find anything on the RFC's that says otherwise.
Also, it's handy to be able to deal with these special addresses - despite the fact that they're not valid in the sense that they can't be assigned to an interface.

[ioprev]

Sure, its pretty useful when handling route addresses. Thank you.

[msimkins]

As an address it would be ::/128 - and that is defined as the 'Unspecified Address'

[ioprev]

That's correct. ::/128 is indeed the unspecified address. The ::/0 address is used as the default unicast route address. See section 2.11 of RFC 5156.

[msimkins]

I agree with the statement, but is is an 'Address', as it cannot be assigned to / configured on an interface.

It probably makes no difference, and maybe should be included for completeness.

[jjmanzer]

I would like to poke this bug and get some attention to it (if it has not been fixed already). I have a Cisco config snippit like this....

1090   ip route 0.0.0.0/0 10.30.131.254
1091   ipv6 route 0::/0 xxxx:xxxx:xx:x::1

Now I understand how that looks weird but when a big vendor uses /0 on an IP (4 or 6) one might as well consider it a standard. Can we get this fixed?

[francisluong]

@jjmanzer: I can take a look at this. Can you help me understand more about what the use case is for supporting this? There are operations that IPAddress does that will make no sense whatsoever on a 0 address... e.g. hosts, broadcast, etc...

What are the functions we are interested in? netmask?

[bluemonk]

Sorry, my bad. I said I would have taken care of it but it completely
slipped my mind. I'll give it a spin this weekend. @jjmanzer would you mind
writing a couple of tests just to make sure the bug fix meets your needs?
That'll be extremely helpful.

@francisluong I'll need your precious help to review my rusty ruby before
we release :)
On Jun 28, 2016 17:14, "Francis Luong (Franco)" notifications@github.com
wrote:

@jjmanzer https://github.com/jjmanzer: I can take a look at this. Can
you help me understand more about what the use case is for supporting this?
There are operations that IPAddress does that will make no sense whatsoever
on a 0 address... e.g. hosts, broadcast, etc...

What are the functions we are interested in? netmask?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#38 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AAFW7SmmqxFoqE4CJGisDPdyj2dsmDWrks5qQToygaJpZM4B-i-B
.

[francisluong]

@bluemonk I'm in!

[jjmanzer]

So a quick reason on why you would see this use case is that 0.0.0.0/0 and 0::/0 are used to say 'match every possible route'. You would not see this kind of IP bound to a network interface but instead see it elsewhere in a device's running config somewhere near the routing directives.

So not every call to the ipaddress gem is for an actual host IP, sometimes it is for less common things like this.

For example, we could have a filter that allows any route that is /32 only.
ip prefix-list TEST seq 10 permit 0.0.0.0/0 ge 32