Re: [netmod] question regarding IPv6 address format / canonical

[Mikael Abrahamsson <[e-mail]>]

On Fri, 7 Dec 2018, Ladislav Lhotka wrote:

> Hi Mikael,
>
> On Fri, 2018-12-07 at 09:20 +0100, Mikael Abrahamsson wrote:
>> Hi,
>>
>> we've had an interesting interop problem, and we don't know if this is a
>> client, server, or just interop problem. However, I thought I'd bring it
>> to the attention here.
>>
>> The server produced an output that was in the format of:
>>
>> 2a00:db8:1:2:3::5:0
>
> Correct - this is the canonical value.

"Single all-zero 16-bit chunks are not compressed. "

So that doesn't seem like the canonical value to me?

>>
>> When the client then asked for information about this object it used:
>>
>> 2a00:db8:1:2:3:0:5:0

This is the canonical value according to my interpretation.

>> The netconf server then returned no answer, because it didn't consider
>> these to be the same (string match).
>
> This is a bug in server implementation.

Can you please elaborate? What is the bug here?

>>
>> I have included what I think is relevant text below, it seems the client
>> reformatted the address into canonical format. However, the description
>> below seems to indicate that all those IPv6 types are ok. If the server
>> must use canonical format, is there a MUST somewhere that says so?
>>
>> What does it mean that something is a "canonical format"?
>
> It is the value that the server (conceptually) uses internally. See sec. 9.1 in
> RFC 7950.

Ok, so if the server doesn't re-format and always uses string in the 
canonical format, that's a clear bug?

-- 
Mikael Abrahamsson    email: [e-mail]