The early RFCs for DHCP stated that the DHCP server MUST support bootp.
It seems to be a common misconception that this is still true. The
current references state that it MAY support bootp. This means that it is
an optional item, although it is desirable. As such, Microsoft's
implementation is conforming to the published specifications (possibly
with 'bugs', but then any non-trivial program will have bugs).
On the subject of the ICMP, I do agree with you. However, it is not fair
to say that it is a design consideration of Microsoft's TCP/IP
implementation. It is simply a BUG. I would hate to have made such a
statement about Linux's TCP/IP code having based it upon Net-2 (the
predecessor of Net-2-Debugged, ...)! There were enough bugs in that
implementation. It brought down enough networks. So, please don't
overreact about bugs. Let alone make comments as you did. There probably
are still bugs in the current TCP/IP code for Linux. Does the fact that
there are bugs mean that it is not conforming to specifications? Perhaps
. . . . Even BSD networking does not conform to the internet RFCs where
it suits them not to do so.
I agree with Alan. Complain to Microsoft about their bug in their
implementation.
There are proper procedures for reporting the problems to them so that
they can track the problem and have it fixed. That is the proper way to
have the issue addressed. Saying that this is a bug to the Linux
developer's list and somehow hoping that it will magically get into the
bug database for Microsoft, or, as your statement "that big companies
don't obey the standards even when they complain to implement one" [sic.]
was not appropriate.
I am sorry to take this message to the list. I am not trying to start a
debate/conversation. However, my 'complaint' was originally mailed to
Ulrich directly without being cc'ed here. His reply was sent here so this
is my rebuttal. Please, if you do disagree with my note, send me private
email. This list is already terribly lagged.
----------
From: Ulrich Windl[SMTP:Ulrich.Windl@rz.uni-regensburg.de]
Sent: Friday, April 12, 1996 8:25 AM
To: Al Longyear
Cc: linux-kernel
Subject: Re: firewall, reject: icmp vs. tcp
It seems a lot of people did not like my statement quoted below.
Therefore I'd like to say what's below.
On 10 Apr 96 at 13:43, Al Longyear wrote:
> You are entitled to your opinions. However, don't let things like facts
> diswade you from your beliefs.
>
> (You are incorrect in your statements as neither is true.)
I'm absolutely sure that the Win95 DHCP client is against the specs
in RFC1541. I've debugged a DHCP server with Win95; I know. Please
try to send your Win95 a hostname via DHCP, or examine the timeouts
when Win95 is retrying.
For IP I think ICMP is an integral, not optional part of IP.
Therefore if "destination is unreachable" is not treated correctly,
it's not IP what they implemented. (I hope I remembered it reight --
we had a 24 hour EMail breakdown twice this week)
>
> > It really makes me angry that big companies don't obey the standards
> > even when they complain to implement one. For eyample Win95 does NOT
> > implement DHCP correctly; as Alan indicates they didn't implement IP
> > correctly, too. But people buy these things and then they complain...
PLEASE: Don't insult me; try to discuss instead.
Excuse me! That was **YOUR** paragraph! Didn't you recognize your own
words? If you find it insulting, then consider how everyone else felt
receiving those comments FROM YOU.