NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/sys/hp/14372

Text File | 1992-12-25 | 2.9 KB | 51 lines

Newsgroups: comp.sys.hp
Path: sparky!uunet!paladin.american.edu!gatech!destroyer!cs.ubc.ca!utcsri!helios.physics.utoronto.ca!alchemy.chem.utoronto.ca!system
From: [e-mail] (System Admin (Mike Peterson))
Subject: Re: Problems with 'named'
Message-ID: <[e-mail]>
Organization: University of Toronto Chemistry Department
References: <[e-mail]> <[e-mail]>
Date: Fri, 25 Dec 1992 14:28:41 GMT
Lines: 40

In article <[e-mail]> [e-mail] (Art Harkin) writes:
>/ comp.sys.hp / [e-mail] (System Admin (Mike Peterson)) /  9:40 am  Dec 15, 1992 /
>
>> Unless your /etc/resolv.conf file contains nonsense, or you are
>> resolving names by NIS instead of DNS, you may have fallen on the
>> $INCLUDE bug in HP-UX named --- if you use $INCLUDE, it will destroy
>> your named data, since things like SOA records in the included file
>> are not seen, and IP address, etc, are garbled (many of ours got 255
>> inserted into one of the fields, which were easy to spot from a database
>> dump).
>
>  This is not entirely correct. $INCLUDE files do work, with the exception
>of one known bug (originally in BSD), that does not correctly count SOA RRs
>within a zone. This bug only takes affect if the SOA record is located in
>the INCLUDE file rather than the first level database file. In this particular
>case the database (after being loaded) is considered erroneous, and is not
>considered authoratative. The data then can be replaced by garbled data that
>looks authorative which may be the case for Utoronto.

I hate to be picky, but what I said is precisely correct, and is in fact
exactly what you just said in your posting! Using $INCLUDE causes the
SOA record to be ignored, and the named data becomes corrupted.

I have posted about this problem several times in response to other
people asking why named doesn't work, and in those postings I have also
stated that the problem also exists on HP Domain/OS systems, and thus
is probably in the original source code. The version of named that I
run has had some patches applied to correct the problem, which hits us
severely because we must run dual domains (".ca" and ".edu") to be able
to talk to our campus and the whole world at the same time; this is most
easily done by INCLUDEing the host data with different domains to avoid
editting two copies of nearly identical files. We also run redundant
master nameds, which have identical data (except for the SOA records).

Given that I originally called in this problem in June, that it was not
fixed in 9.0 and that a patch is still not available is none too impressive.
--
core error - bus dumped    -*-    Mike Peterson, SysAdmin, U/Toronto Chemistry
 *******   As usual, I speak only for me, myself and I; nobody else   *******
E-mail: [e-mail]  Tel: [phone]  Fax: [phone]