I'm giving this one last attempt.
Quality Quorum wrote:
On Fri, 21 Mar 2003, Greg Daley wrote:
I'm pretty sure this would cripple any (reasonable) end-to-end assumptions people may want to make with programming IPv6 apps.
I do not think so.
Well, have you ever seen the hoops required for a NAT to undertake H.323 or another peer to peer application with multiple channels? NAT only works relatively easily with single session client apps.
P-to-P applications are a candidate NextBigThing.
Of course, we can't support NAT modifications for every application which will be programmed, so devices on NAT'ed networks will not be able to participate in p-to-p.
IPv4 networks are rife with NAT, so IPv6 adoption may be driven by p-to-p. If NAT is good enough, there is currently no significant reason to go to IPv6.
Please, think of your children when asking us to support NAT. We're trying to make the Internet a system which will support the applications we need in 2020, not 1999.
If you are making assumptions that there will be no NAT6, but it may be forced on us by the market, it is you who are crippling the Internet of the future.
Actually, I'm not. I'm asking that all efforts toward NAT6 be dropped until there is interest from the user community. We know how NAT4 works, NAT6 will be trivial.
In the mean time we'll try to wean people off NAT, by providing better applications and connectivity with end-to-end IPv6.
I'm not advocating "fight against the tide", but initially IPv6 can be marketed as better than IPv4 because of NAT4. Once IPv6 starts taking off, then it's our job to educate people about NAT being evil.
If you are successful there will be no NAT6 even it if it is supported, however if you are unsuccessful the damage done by the no-NAT6 decision will be significant.
There is no damage to be done. We've got almost 2^61 networks to give away, and almost 2^45 organizations to give addresses to. It will be at least 3 years before there are 2^45 people in the world :)
NAT can come in if we get to the point when we've used up 2^44 organizational allocations and have 2^44 left. I'd suggest that by that time we'd be able to go to the next IPng though.
We shouldn't be accomodating it now.
Your own argument points the other way - if we are thinking about future we should not artificially cripple the environment now.
End-to-end will never cripple NAT. NAT can ALWAYS be retrofitted to an end-to-end network (maybe some applications won't work, but we know ones that will). If there's need, we can do it simply.
Running end-to-end applications on a NAT network is a nightmare, though.
Why don't we wait for the problem to arise before we implement NAT? It's not like we don't have better things to do.
Greg
-------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------