Hi,

   I have just freshly installed redhat 7.1 re i386 iso
   I've used linux on and off since 91 or 92.

   xinetd has a naughty ftpd problem which [is not] yet 
   in any problem report.

   xinetd's man page states, for "/etc/xinetd.d/*",
   if disable == yes for a service
   that xinetd, when a request for that service on
   the port arrives, WILL REFUSE the connection.
   
   --------------------------------
   HOWEVER - xinetd DOES START ftpd on port 23
   despite its flatulent claim!
   (no, no ftpd daemon is or was loaded)
   --------------------------------

I have read the man page.  It is wrong: that is 
an error that needs attention.

xinetd shouldn't launch services on ports it has 
"no configured knowledge of"; as per manpage exposition.

Infact - inetd (Linux NetKit 0.17) has been tried and 
tested.  However:

xinetd has had at least 5 fixes already. xinetd wasn't
initially linuxconf ready (and so valuable linuxconf
contributor time was spent 'fixing it):

I'm wondering just how xinetd got chosen as a viable
unix base system solution.  It just isn't all that:
the author didn't even claim it to be.

Why the name change that breaks scripts (that steer 
inetd)?  Why change the well known super-server fields
in the config file - which up until now have been similar 
across platforms?

Will the linuxconf "fix" that made \[Xi]netd actually work
keep me from using the inetd from:
        "Linux NetKit 0.17"
with redhat?  If it does - I'm telling!  (Or, I'll just not
be as happy as I would normally be.)

Thanks, hope that helps,

    John D. Hendrickson

jdh@hend.net

johndhendrickson22124@yahoo.com


P.S

   This is like strike six for xinetd, right?  How many
   strikes does it get anyway?  

   Is an imperfect kerberos implementation really important?  

   Kerberos doesn't have three heads -- it has 100 
   configuration sides.  I mean, an NIS+ configured subnet 
   with stunnel-only connects is good too -- and offers the 
   same opertunity for password upgrade binges, host-host 
   tunneling, secure administration, etc...

   And anyway?  If kerberos is so great, how come they're
   "begging" it be exported?

   Ok - maybe I'm out of my realm on the encrypted 
   configurable network point.  Still - stunnel is easier
   and doesn't require application re-writes for transport
   independant (unix) apps, and Kerberos does (kerberiz#q*^&)  
   
   Right?

   Anyway, Have a good one..
      jdh