[PATCH RFC v2 0/3] dhclient: add --prefix-len option

Shawn Routhier sar at isc.org
Wed Dec 2 20:20:37 UTC 2015

> On Dec 2, 2015, at 11:25 AM, H. Peter Anvin <hpa at zytor.com> wrote:
> On 12/02/15 11:21, H. Peter Anvin wrote:
>> I have the following questions with regards to the protocol and
>> desired behavior:
>> - The current patch will only emit a null prefix for a SOLICIT
>>  message, but are there other messages where a null prefix is also
>>  appropriate?
>> - The current patch specifies a global prefix length, but would it be
>>  better if that prefix length was settable for each IA_PD IAID?  In
>>  that case, what do people think of any of the following command line
>>  syntaxes (where # is the number of bits):
>>  -P#
>>  -P=#
>>  --prefix=#
>>  -P #
>>  The last one would break if it is possible for a purely numeric
>>  argument to follow a -P option on the current command line (I don't
>>  know if there are any systems that use numbers for interface names,
>>  perhaps?)
> One final question: why is the t2 parameter of the IAADDR and IAPREFIX
> option set to t1 + (t1 / 2) for SOLICIT, mirroring the IA_NA/TA/PD
> option, but t1 + 300 for REQUEST, RENEW, and REBIND?  I have kept that
> behavior in this set as I wanted the first patch to have no functional
> changes at all; if implemented properly it should behave exactly like
> the old code.
> 	-hpa

Thank you for your work on this feature.   I’ll have to think about your questions
and am busy with other tasks right now so I’m not sure when I will have the time.

One item I did want to bring to your attention is that I have just recently merged
a patch to handle getting IA_NAs and IA_PDs in the same message better.  This
is from RFC7550 and can be seen in the sources from source.is.org.  I mention
this as this patch also changed the dhclient code and overlaps the changes you
are proposing in your patch.

Shawn Routhier

More information about the dhcp-workers mailing list