Hi Theo,

On 3/7/21 5:45 PM, Theo de Raadt wrote:
> Paul Kelly <[email protected]> wrote:
> 
>> So...here is a proposal for updating various man pages to try help
>> others who make a similar migration to dhcpleased. It's a bit of a
>> scattershot approach and the wording can probably be improved, but it is
>> hopefully consistent with the existing pattern for describing auto
>> interface configuration, and it at least should prompt a user to
>> double-check their interface state before they (incorrectly) conclude
>> that dhcpleased doesn't work on trunk interfaces.
> 
> I don't understand where you are going with this type of documentation.

Fair enough. On reflection, the change to dhcpleased.8 isn't useful
because it clumsily tries to make a point about dhcpleased behaviour
that is only relevant in the context of dhclient behaviour.

I still think the changes for hostname.if and ifconfig are relevant in
the context of those man pages as they currently stand, because dhclient
and slaacd are already described there. I'll try to add some
justification for that below, but I also note that you're hinting at
bigger things to come so it might all be redundant in the big picture.


> Of course the interface has to be up.  Interfaces that are down don't
> participate in networking.  It is obvious.

It's a fair point, and it's certainly obvious when not relying on
dhclient to do it automatically. It wasn't clear for me in the
transition from dhclient to dhcpleased that the behaviour was different
in this respect, and the commit message for dhcpleased seemed to
indicate that it should only require a substitution of "dhcp" -> "inet
autoconf" in hostname.if. The implicit reliance on dhclient's automatic
behaviour is also present in the faq - faq6 makes a point that "dhcp" is
all that is required with dhclient. So I wouldn't be too surprised if
there are a number of people out there who could try to make the
transition and end up scratching their head like I did.

In my view there is also a small disconnect between the man pages for
dhcpleased man page and hostname.if. The dhcpleased man page mentions
the AUTOCONF4 flag and then delegates the details to the man pages for
hostname.if and ifconfig. Those man pages don't appear to mention
dhcpleased at all, and instead they provide context and examples for
dhclient and slaacd. This is a bit jarring and leaves the impression
that something is missing.

So I don't mean to say there are major faults in the documentation here;
but I do think that these relatively small changes would make the
transition from dhclient more obvious.


> We are discussing bringing the interfaces up in a more automatic
> fashion, and then this won't matter.

I can't really know exactly what you're alluding to here, but it's
certainly possible to see some trends relating to network configuration
in the past few weeks. I'm certainly interested to see where it goes. My
presumption would normally be that this kind of work would be delayed
until 6.10, based on the current timing of 6.9. But maybe there are
still some surprises in store for the 6.9 release as well :)

Reply via email to