From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <6e7c4f29aeb63684583d88faa52cc7d4@plan9.bell-labs.com>
From: presotto@plan9.bell-labs.com
To: 9fans@cse.psu.edu
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: [9fans] new stuff in sources.
Date: Sat, 15 Jun 2002 20:46:20 -0400
Topicbox-Message-UUID: aeedb13e-eaca-11e9-9e20-41e7f4b1d025

I've update libndb and dns on sources.cs.bell-labs.com.

libndb:

1) I made the number of cached search responses larger.  The caches
  are flushed if the files they represent are changed.

2) I took out the need for 'ipsubmask=' entries in /lib/ndb.
  It used to be necessary for ndbiplookup.  This routine
  looks for attributes accociated with an IP address.  If
  it doesn't find the attribute in the entry for the system,
  it looks in the entries of subnets that contain that system.

  The search used to work by starting at the class A, B, or C
  net and working down through successive subnets, steared by
  the ipsubmask entries.

  Now we just start with the address and peel back a bit at
  a time (starting at the lowest order bit) looking for a subnet
  entry.

  That means you can now just declare a subnet like:

	ipnet=quux ip=10.1.1.0 ipmask=255.255.255.0

  without having to also declare

	ipnet=classA ip-10.0.0.0 ipsubmask=255.255.255.0

  It should make life easier.  It's definitely easier to
  explain since noone knows what Class A, B, and C IP nets
  are now that the world is all CIDR.

  The potential down side is more searching through the
  database, hence the larger number of cached responses.

dns:

1) We used to create a PTR entry for every IP address.
  We now do it only if there is an SOA record for the
  subdomain.  For example, if we have an soa like

	dom=9.104.135.in-addr.arpa soa=
		ns=ns1.cs.bell-labs.com
		ns=ns2.cs.bell-labs.com

  then an entry like

	dom=achille.cs.bell-labs.com ip=135.104.9.2

  will automaticly generate a pointer entry like

	dom=2.9.104.135.in-addr.arpa
		ptr=achille.cs.bell-labs.com

  There's an added hack that allows you to generate PTR's
  ala rfc2317.  That RFC tries to deal with delegating
  reverse lookups for network masks that don't end on a
  byte boundary.  It only works for masks that end in the
  lowest order address byte.  It does let ISP's like
  Verizon to hand out a handful of addresses to customers
  and delegate reverse lookups to the customers.

  The format of the reverse entry is not completely
  defined.  However we'll accept both forms suggested
  in the RFC.  For example, if we have a subnet
  defined as:

	ipnet=quux ip=135.104.8.0 ipmask=255.255.255.128

  and also the soa

	dom=0.8.104.135.in-addr.arpa soa=
		ns=ns1.cs.bell-labs.com
		ns=ns2.cs.bell-labs.com

  the entry

	dom=trex5.cs.bell-labs.com ip=135.105.8.1

  would also generate a pointer entry

	dom=1.0.8.104.135.in-addr.arpa
		ptr=trex5.cs.bell-labs.com

-------------

We've been running this for a few weeks on our systems.
Complaints/problems to me.