Re: NT 3.51 dhcp client and server ip address

[Shawn Mamros <[e-mail]>]

[e-mail] (Don Coolidge) writes:
>Leaving it as is simply postpones fixing Ted's problem, yes?  And while
>fixing it would require reworking of some existing implementations, that
>would have to be done anyway when the problem is finally dealt with, as it
>must be eventually. Fixing it now frontloads that effort and leads more
>quickly to common interpretation of the spec and greater interoperability.

As Mike Carney and I have both already said, there are practical and
*already implemented* means by which multihomed DHCP servers can be
made to work, while leaving the current definition of the Server
Identifier option intact.

Thus, there is absolutely NO REASON to be making such a *fundamental*
change to the DHCPv4 protocol at this point, and to cause needless
grief to all of the vendors who have working software out there *today*
in the hands of paying customers.

And ultimately, at the end of the day, it is those customers who will
have the final say on what protocol changes do and don't get accepted.
DHCPv4 has LONG since passed the point of experimentation - there are
too many people out there who rely on it day-to-day.

I'm not saying that DHCPv4 is a perfect protocol.  It has plenty of
warts, as many protocols (all the way down to IPv4 itself) do.  The
best thing we can do is document those warts as thoroughly as possible,
so as to let everyone know what the "commonly accepted practice" is.
If that means the Draft Standard gets delayed, so be it.  The marketplace
has already passed far ahead of that point.

(If I were a politically-minded person, I could potentially say that
perhaps your company affiliation might be motivating you to be doing
this, so as to potentially inconvenience a large competitor of yours
located several hundred miles to the north of you.  But I'm not so
politically motivated - except when I have to be. :-)

-Shawn Mamros
E-mail to: [e-mail]
(Opinions expressed are my own, not necessarily those of my employer)