Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!hplabs!hp-ses!hpcuhb!hpindda!dfc
From: dfc@hpindda.HP.COM (Don Coolidge)
Newsgroups: comp.sys.hp
Subject: Re: BROADCAST ADDRESS ON HP-UX.
Message-ID: <4310103@hpindda.HP.COM>
Date: 19 Mar 90 23:14:17 GMT
References: <608@inesc.UUCP>
Organization: HP Information Networks, Cupertino, CA
Lines: 48

>What a crock of arrogance!  Just about every "modern" TCP/IP implementation
>acknowledges that there will be pre-standard systems on the net, and
>permits setting of broadcast addresses.  Why does HP think it "knows better"
>than the users? 

Ahem...

Yes, it can be interpreted somewhat arrogantly (though not intended that way).
Unfortunately, so can your response. Care to think it over a bit, and perhaps
post in a somewhat calmer manner?

For the record (as someone who actually does kernel networking code 
development in that area) [ oh, yes - for _my_ record, not necessarily
legally HP's ]:

	1) The current HP-UX networking code is based on 4.2BSD, not 4.3BSD.
 	   The ability to arbitrarily configure your interface's broadcast 
	   address is a standard part of the 4.3 networking release, but not of
	   4.2. If by "modern" you mean "4.3", then you're correct. By that 
	   standard, HP's networking code is not yet "modern" in that it does
	   not yet have all of the 4.3BSD features (nevertheless, it has many).
	   It is, however, of considerably higher quality (less bugs) than are
	   most 4.3BSD products. (Also, 4.3BSD is about to suddenly become no
	   longer "modern" with the release of 4.4BSD) 

	2) Patches are available for the 6.5 and 7.0 (both Series 300 and 800)
	   HP-UX releases that provide that capability. Users wanted it, so
	   we provide it. Contact your HP rep for more details, as the first
	   respondent suggested.

	3) The problem noted by the original poster is one that can only happen
	   in the presence of even less "modern" networking implementations -
	   those based purely on 4.2BSD, with no 4.3 enhancements. HP-UX sends
	   ones (the standard), but also recognizes zeroes. All those other 
	   braindead implementations are *broken*, and should be replaced,
	   regardless of what HP, or any other vendor, does - they violate the
	   Internet standard for broadcasts. They also almost certainly don't
	   have SlowStart, or congestion avoidance, or fast retransmit with
	   exponential backoff, or any of the other good-citizen VJ-Karels/TCP
	   algorithms, all of which HP has (starting in 7.0).

HP most definitely doesn't think it knows better than the users. We listen to
what they have to say. In response to their requests, we're adding more and
more 4.3BSD features with each release, and are tracking 4.4BSD to see what
happens there. We're not all the way there yet, but we're on our way, and 
getting better (and more "modern" ;-) )with each release.

- Don Coolidge