SWIP netblocks

[Gogulski, Mike]

Ginny,

I'd say "no", and provide one example to support my position.

Many ISPs are deploying DSL services using bridged networking models.
Typically, the ISP allocates a large block (/24 or so) at a time to an
integrated routing and bridging interface, and then assigns customer
addresses from this block one at a time as customers come online.  If a
customer requires more than one IP address, these can be assigned, and they
do not have to be assigned on CIDR boundaries at all.

If the assignment is larger than a /29, ARIN policy requires that the
reassignment be documented via SWIP or RWhois.  Implementing this policy
would force an ISP who needed to assign a customer (for example) 9 IP
addresses for a bridged DSL service 16 addresses instead, a net waste of 7
addresses.  This would be in conflict with one of ARIN's other stated policy
goals, namely the conservation of IPv4 address space.

Peace,
Mike Gogulski
Chief Engineer
DSL.net, Inc.

-----Original Message-----
From: ginny listman [mailto:ginny at arin.net]
Sent: Wednesday, January 03, 2001 11:10 AM
To: dbwg at arin.net
Subject: SWIP netblocks


In reviewing what is currently stored in the database, there are a number
of SWIPed netblocks that are not on the bit boundary.  For example,
instead of SWIPing 0 to 255, an entire /24, 1 to 254 was SWIPed.  In the
future, we will be operating in a cidr world, including displaying cidr
blocks in whois.  For a block that is 1 to 254, the display will include 2
/32, 2 /31, 2 /30, 2 /29, 2 /28, 2 /27, and 2 /26.  It would be a whole
lot cleaner to display 1 /24.

How do people feel about enforcing allocations/assignments based on a
single cidr block?  I could see an occasion where someone may want to
assign 2-4 cidr blocks at a single time, but can we enforce, or strongly
encouraging, a policy like this?  SWIP on the bit boundary.

Ginny