[clug] IPv6

Hugh Fisher Hugh.Fisher at anu.edu.au
Sun Jan 16 15:44:46 MST 2011


----- Original Message -----
From: David Austin <david at d-austin.net>
Date: Monday, January 17, 2011 9:17 am


> The IPv4 packet header is nicely generalisable, with versioning 
> and more.
> The IPv4 packet header is at least 20 bytes and up to 60 bytes 
> long (with
> a variable length option field of up to 40 bytes)
> http://en.wikipedia.org/wiki/IPv4#Header.
> So bolting on 96bits/8 bytes is easy.  The trick would be 
> to convince IPv4
> hosts to return the special extended-IP-address-option field when
> replying to packets.  In the decade plus timeframe that 
> we're talking about,
> that should be doable.  Certainly for "important" IPv4 
> sites (web,smtp,etc)
> that'd be quite quick.  Yep it's a kludge.  Perhaps it 
> should use the IPv4
> evil bit to make that clear (http://tools.ietf.org/html/rfc3514).


Guys, all this was discussed - at length, over and over - on the IETF
mailing lists. (I lurked for a while out of interest.) If you want to find
out why certain things in IP6 are the way they are, you could always
go and look in the archives.


I'm not saying that the IP6 development process worked in an ideal
fashion or produced the best result, but was a bunch of smart people
in an open source style environment. 


> Dr Austin is out of luck, but not for technical reasons.  
> IPv6 is the path
> "forward".  I note that the first author of the IPv6 RFC 
> works/worked at
> Cisco
> and Cisco stands to proft the most from IPv6 but perhaps I'm too 
> cynical in
> my old age.


OK, that's low. If you want your network protocol to be developed
by people with expertise and experience in real world networks,
where are they most likely going to be working? This is like complaining
that the PowerPC Linux kernel developers mostly work for IBM.


cheers,
Hugh




More information about the linux mailing list