On Wed, 15 Aug 2007 15:13:06 +0200 Iljitsch van Beijnum <[EMAIL PROTECTED]> wrote:
> On 15-aug-2007, at 12:15, Mark Smith wrote: > > >> I disagree. For better or for worse, the notion of a subnet mask > >> going along with an interface address is deeply ingrained in the way > >> IP is implemented. Separating the two for no apparent reason is a bad > >> idea. > > > I think you need to modify the following sentence as such: > > > "For better or for worse, the notion of a subnet mask > > going along with an interface address is deeply ingrained in the way > > **IPv4** is implemented." > > No, I was very careful about what I said. The only place that I can > think of where you don't configure a mask along with an address is in > the case of HSRP. EVERYWHERE else that I know of you need to specify > it now that we have CIDR and obviously it was implied before we had > CIDR. Yes, in IPv6, too. > > We can talk long and hard about the pros and cons of this > architecturally, but the simple truth is that changing this doesn't > appear to buy as anything tangible and it will require huge amounts > of work in implementations and then generate huge amounts of > confusion in operations so PLEASE let's not go down this road. > I'm don't think I'm going down it. You seemed to be questioning why subnets were fixed length, I provided most likelyreasons why, including evidence that the original design of IPv4, pre-classes, also followed this model. I think that shows that when there is the opportunity to have fixed length node addressing it is very desirable, and has been chosen by many protocol designers (DECNET Phase IV has it too, IIRC) I like that IPv6 can place the boundary between the network and node portion anywhere within the address, unlike IPX and Appletalk. It future proofs the protocol, ensuring that should the network or node portions not be big enough they can be changed, and devices will be able to cope with it without software upgrades. OTOH, I don't like _unnecessary_ complexity. Currently, we don't need to engage any of the complexity of non-fixed network and node portion boundaries in IPv6, so let's not until we need to. Allowing for something doesn't mean you must use it. Simplicity is what we get when we eliminate/avoid/choose not to use variable length subnets, and it is a very tangible property, well worth pursuing. I could probably save 5 minutes per day at work if I didn't having to think through or calculate variable length subnet masks in IPv4. That's 1200 minutes or 20 hours a year. Multiply that by how many operators there are out there spending that 5 minutes or more per day and it's a huge amount of spent time. That's a necessary expense when dealing with IPv4, but not a necessary one for IPv6. If it's not necessary, then it's wasted time. > And if you have to look back before RFC 791 to provide > counterexamples that really doesn't disprove "ingrained" in my > opinion... That's probably a fair critism. I was mainly replying to what I've realised was an earlier question of yours, regarding why fixed length subnets. The ingrained bit seemed to me to suggesting that fixed length subnets would be considered to be a radical idea and wouldn't be acceptable to those that have led sheltered protocol lives and only ever dealt with IPv4, post 1995. Regards, Mark. -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
