Subject: Re: NetBSD-3 NIS-compat getpwnam()/getpwuid iterate entire map [was Re: 3.0 YP lookup latency]
To: Hubert Feyrer <hubert@feyrer.de>
From: None <jonathan@dsg.stanford.edu>
List: tech-net
Date: 06/21/2006 16:57:31
In message <Pine.LNX.4.61.0606220052550.7825@m24s24.vlinux.de>Hubert Feyrer writes
>On Wed, 21 Jun 2006 jonathan@dsg.stanford.edu wrote:
>> + if (strcmp(nconf->nc_proto, "tcp", (size_t)3) == 0)
>
>strcmp only has two args... is there a "tcp6" value to be expected?
Nope, It's's just me being even more stupid than usual, due to a sinus
infection. Thanks for catching it. The version I compiled to test
before sending still had the n in str[n]cmp(). I think both the 'n'
and the lenght argument should go. My apologies to you, as well as
Charles.
netconfig(5) does not contstrain "network_id" (the purpose of
netconfig(5) is to _define_ usable network_ids!). The examples therein
define "tcp6" and "udp6". netconfig(5) does explicitly say that the
"netproto" field must be one of "tcp", "udp", or empty. Bizarre,
given that the raison d'e^tre of TI-RPC is to be
"transport-independent"; but that's where hype meets reality.
I doubt anyone, anywhere, is using TCP over non-IP(v4,v6) transport.
Or at least not with NetBSD. Jason removed TUBA from our tree nearly
10 years ago. Non-TCP, ordered-bytestream protocols: NETISO has TP4.
Did anyone ever help Randy Stewart to finish integrating SCTP into NetBSD?
I see Solaris 9 requires non-zero-length strings, "tcp", "udp" or
"icmp", with "-" to indicate ``none of the above''. according to
http://www.opensolaris.org/jive/thread.jspa?threadID=9766&tstart=0
TI-RPC on Solaris cannot run over SCTP, because the Solaris SCTP
implementors focused on the BSD sockets API, whereas the Solaris
TI-RPC implementation uses the SysV STREAMS API. Ironic.
Does anyone actually use TP0/TP4? Or want it badly enough to fix our
(TI-)RPC to actually work over TP4? Or to test an attempt to make that work??