[OpenIndiana-discuss] How to configure a DUID for DHCPv6 usage?

[Henner Gratz]

On 04/21/2016 01:37 PM, James Carlson wrote:
> On 04/21/16 06:52, Henner Gratz wrote:
>> Hello again!
>>
>> In the meantime I've worked through states.c and found a few bugs that
>> make the correct
>> parsing of DUID-LLT's and DUID-LL's impossible. A self-compiled,
>> corrected version of
>> dhcpagent seems to work (at least for me ;*). I filed a bug
>> (https://www.illumos.org/issues/6928) with a diff-file. Hope this will
>> help a little. (I run OI_151a9, but
>> the source of dhcpagent is still the same in the current illumos sources.)
> As I've tried to explain several times over, the time value in the type
> 1 DUID was *NOT* intended to be specified by the user.  It shouldn't be.
>   The DUID (once generated) is stored in stable storage and reused, so if
> that's the reason you're trying to specify it manually, then you
> shouldn't do that.
>
> For that reason, I disagree with that part of the change.  I don't
> believe it's correct or necessary at all for stable DUIDs.
>
> I'll leave the rest for a maintainer to look at.  I believe this is a
> fundamental misunderstanding of what the code is doing.
>

Of course, you know much better what once was intended by the Sun 
engineers, but...

I have a development environment where it's necessary/useful to replace 
a machine
with another without having to change the DHCPv6-server's configuration 
as well.
So it's necessary to set an identical DUID on all machines. (I know that 
this is not
what was intended with RFC 3315.) OK, I could take a hex editor and poke 
around in
/etc/dhcp/duid, but this would be an even uglier way. So I use the 
documented way
to assign a certain DUID to a system: Adding a CLIENT-ID entry to 
/etc/default/
dhcpagent. And as the this DUID takes precedence over /etc/dhcp/duid and 
so takes
its function, it should be as stable and persistent as the DUID saved in 
the duid-file.
And I don't see why it shouldn't be possible to set the timestamp of the 
DUID as well.
In the end a DUID is just a bunch of byte-values. (That normally should 
be unique
in a network. - I think, I understand RFC 3315 and its intention in this 
regard quite
well.)


The timestamp in type 1 DUID's is only there to make it even less likely 
that
identical DUID's are created by automatic processes. But entries in 
/etc/default
/dhcpagent are the way for administrators to configure the DUID by hand 
- if necessary.
And admins should know what they do. BTW, identical DUID's aren't a 
problem in a network
as long as only of these machines is connected to it. But for 
hot-stand-by solutions or -
as in my case - in development environments, identical DUID's can be a 
good thing.
That's my 2 cents. And know I will shut up and wait for the verdict of 
the maintainers.



Regards,

     Henner