[Samba] Again kea DHCP-Server

[Gregory Sloop]

> Well, I have a slightly different opinion.

> 1. I’ll channel Jon Postel here a little bit — Networks and Systems should be as simple
> as possible and no simpler… IMHO, DNSMASQ is too simple for most of the
> DHCP environments I deal with and I find its structure frustrating at best.

> 2. Yes, KEA seems over complex when you first approach it. Mainly, that’s because
> there’s too much text book documentation and not enough basic example and
> HOWTO out there (yet). KEA is definitely capable of being much more complex
> than traditional ISC DHCPd that we all know and love… However…

> 3. ISC DHCPd has become quite a hodgepodge of hacks and a conglomeration
> of code spaghetti over the years as various ornaments got hung on the DHCP
> tree. KEA is a concerted effort by ISC to step back and look at creating a clean
> server with all the modern capabilities needed in a reference implementation
> of DHCP that provides literally every possible feature. IMHO, it’s an excellent
> effort by very talented people.

> That said, it does have some shortcomings. ARM support has been limited
> (nonexistent) until very recently. ARM packages still aren’t out (yet), but there
> is good progress towards this recently and I am hopeful that we will see full
> ARM support for KEA on par with x86 very soon now.

> In most use cases, KEA (once you get past a small learning curve) isn’t significantly
> harder to manage than ISC DHCP and actually offers much greater flexibility in where
> you put various things in the configuration file and how you manage things like
> reservations, pools, options, etc.

> The Client Classing engine in Kea is top notch and very functional, but again, it
> does come with a bit of a learning curve. The ability to have separate namespaces
> for vendor options (and custom options) will appeal to anyone who has had to
> deal with subnets with more than one different and incompatible vendor-specific
> use of DHCP options 43 and 60. (try that with DNSMASQ… I dare you).

> Unfortunately, i lack the Samba experience to reliably implement what is needed
> here, but I am happy to provide what kea expertise and experience I have to an
> effort to address this issue if someone more versed in Samba wants to collaborate.

> Owen

 
All this, AND
ISC's dhcpd is now deprecated and all new features etc will only see developer time for Kea. 
dhcpd is on life-support.
 
Sure, it'll probably get patches for a good while yet. There will probably be people still using it for another dozen years or more. But Kea is where sysadmins with forward-looking stances will look to migrate. (And, having been victim of a few really odd bugs in dhcpd [especially in embedded versions] I have hopes that Kea will be better.]
 
The next time I have any major work on my largest pair, I'm migrating to Kea - and I've already started looking at the details, setting up a lab-environment server to test on etc.
 
So, yeah - Kea is def where I'm headed for anything other than really trivial/small setups.  
Having Samba support Kea sooner rather than later would be good.
 
-Greg