> In article <20080127084912.56D7B11703@yamt.dyndns.org>,
> YAMAMOTO Takashi <yamt@mwd.biglobe.ne.jp> wrote:
> >> On Jan 26,  7:57pm, yamt@mwd.biglobe.ne.jp (YAMAMOTO Takashi) wrote:
> >> -- Subject: Re: CVS commit: src/lib/libc/rpc
> >> 
> >> | > Module Name:	src
> >> | > Committed By:	christos
> >> | > Date:		Fri Jan 25 19:43:53 UTC 2008
> >> | > 
> >> | > Modified Files:
> >> | > 	src/lib/libc/rpc: pmap_getport.c
> >> | > 
> >> | > Log Message:
> >> | > PR/37864: Wolfgang Stukenbrock: when requesting TCP rcp-service
> >port numbers
> >> | > UDP is used
> >> | > 
> >> | > 
> >> | > To generate a diff of this commit:
> >> | > cvs rdiff -r1.16 -r1.17 src/lib/libc/rpc/pmap_getport.c
> >> | > 
> >> | > Please note that diffs are not public domain; they are subject to the
> >> | > copyright notices on the relevant files.
> >> | 
> >> | does it make sense to retry on RPC_PROGNOTREGISTERED?
> >> 
> >> Can't a program register only for tcp and not for udp?
> >> 
> >> christos
> >
> >your change is about the transports used by the portmapper itself,
> >not by the queried services, right?
> 
> Yes, I was confused. It does not make sense to retry if the program
> was found not to be registered.
> 
> >honestly speaking, i don't understand the rationale of this change.
> >if tcp is so reliable, why don't you try it when protocol != IPPROTO_TCP?
> 
> I think that Wolfgang was trying to use NIS across a network where UDP
> was not available/reliable.

then, doesn't it make more sense to always try both transports?

YAMAMOTO Takashi

> 
> christos