separating readers and peers

[Bill Davidsen]

Russ Allbery wrote:
>Christian Kratzer <ck-lists at cksoft.de> writes:
>
>  
>
>>thats what I had in mind but I was concernded nnrpd would not handle
>>authentication when started this way.
>>    
>>
>
>innd is just a weird inetd from nnrpd's perspective.  innd doesn't do
>anything for nnrpd that inetd doesn't do.
>
>  
>
>>Do you see any caveats from running nnrpd directly using -b address -D ?
>>    
>>
>
>This will also work; just be aware that nnrpd -D isn't as widely tested
>and has in the past had some strange issues (although they seem to have
>mostly been fixed).
>
>  
>
It's pretty well tested, we run the backend databases that way, with the 
innd feed on one port and the nnrpd daemon running on another. Does 
speed up starting a new connection, since there's no fork/exec pair, 
just a fork. I have to wonder if threads would be helpful here, but from 
other projects I doubt it would be worth the effort to rewrite, and 
certainly less portable.

-- 
bill davidsen <davidsen at tmr.com>
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979