[unisog] Mac OS X 10.4.x "DHCP client sometimes remains BOUND aftersending DHCPDISCOVER" bug

[David W. Hankins]

On Thu, Feb 01, 2007 at 03:52:35PM -0700, Ted Lemon wrote:
> I have to agree here: the behavior ascribed to the OSX 10.4 client is  
> neither in conformance with RFC2131 nor recommended by RFC4436.    
> RFC4436 pretty clearly talks about INIT-REBOOT, not INIT for sending  
> DHCP messages when reconfirming an address for which a valid lease is  
> held.   If you read the glossary definition for Operable Address in  
> the draft, you can see that it is specifically talking about this.    
> I don't really see any room for interpretation here.

Generally I agree, I just disagree that this was the intent,
which is different from the syntactical content.

I think that glossary entry is an oversight.  If the author
really intended this interpretation, he should have just used
'unexpired lease', as that wouldn't require a glossary entry
and is already used and well understood in other documents.

The DHCWG should have told them as much during review.

If 'unexpired' was the author's intent, then he really went out
of his way to create indirection by inventing a new term.

> So I would say that the DHCP server that you are using is not  
> behaving out of spec, at least by my reading of it, although  
> certainly its behavior is more restrictive than usual.

But I have to disagree here.

Reducing expiration times on any active lease except when you
receive a DHCPRELEASE or DHCPDECLINE is patently dangerous (but
ISC DHCP allows it anyway on DHCPREQUEST because sysadmins like to
have that tool).  You can't know that the client will receive and
bind to the server's response packet, it could be lost.

My advice is to only update the lease expiration time if it
moves into the future.

-- 
David W. Hankins	"If you don't do it right the first time,
Software Engineer		you'll just have to do it again."
Internet Systems Consortium, Inc.	-- Jack T. Hankins