>Priscilla Oppenheimer wrote:
>>
>> Zsombor Papp wrote:
>> >
>> > As a side question, do you think that TCP must run over IP? :)
>>
>> I forgot to comment on that very important question! :-)
>>
>> I've never seen TCP run over anything other than IP, although
>> in theory it could.
>
>Has anyone seen TCP run over anything other than IP? I'm curious.
Not as far as I know. XTP replaced both. While it was bastardized
OSI, the MAP Enhanced Performance Architecture ran the application
over LLC Type 3 and had neither transport nor network, but still was
reliable.
>
>> The devil is in the details though.
>>
>> Assuming no changes and compatibility with RFC 793, the RFC
>> that specifies TCP, TCP must run over IPv4. Check out how the
>> TCP checksum works. It's based partly on a pseudo header, which
>> includes fields from the IPv4 header. The pseudo header
>> contains the Source Address, Destination Address, Protocol, and
>> TCP length. This gives TCP protection against misrouted segments.
>>
>> So, you couldn't just plug in some other network layer and
>> expect TCP to work unchanged.
>>
>> So, what are the IPv6 people doing about this? Anyone know?
>
>I was hoping somebody else would research it and save me some work :-), but
>it wasn't too hard to research so that's OK. The RFC for IPv6, (2460)
>addresses the issue. It says:
>
>Any transport or other upper-layer protocol that includes the addresses from
>the IP header in its checksum computation must be modified for use over
>IPv6, to include the 128-bit IPv6 addresses instead of 32-bit IPv4
>addresses. This includes TCP and UDP.
Of course, if you encapsulate IPv4 in IPv6 and run TCP over v4, you
finesse the checksum problem until TCPv6 is available.
>
>And here's an interesting thing: The IPv6 version of ICMP [ICMPv6] will have
>a pseudo header now too. The reason for the change is to protect ICMP from
>misdelivery or corruption of those fields of the IPv6 header on which it
>depends that aren't covered by the checksum. I think this has to do with the
>fact that they moved a bunch of stuff from the main header to extension
>headers.
Sortakinda. ICMPv6, which you have to look at in combination with
Router Discovery and Neighbor Discovery, is pretty much a new
protocol as is IPv6 when compared to IPv4.
At least in V6, ARP goes away, and we won't have the arguments as to
what layer it's in!
>
>
>Anyway, the concept that protocols are so modular that you can plug them in
>anywhere and it will still work is over-simplified. You can plug a serial
>interface encapsulation under anything on Cisco routers, as discussed ad
>nauseum, but that doesn't mean you can do it with other protocols. You can't
>even plug IPv6 in for IPv4 and expect TCP to work without modification!
US Navy wisdom:
1. If it doesn't fit, get a hammer.
2. If it still doesn't fit, get a bigger hammer.
3. If it still doesn't fit, get the submarine-rated duct tape, wrap, and
pray.
Alternatively, collapse Schwarzenegger's second law, and you can
solve an amazing number of things with shim layers and tunneling.
>
>IPv6 should be lots of fun, eh? Doesn't sound as bad as the dire warnings
>from Howard about the current model of BGP and addressing, route churn and
>Internet instability, though!
>
>And on that cheery note, I will leave you for an ice-cold margarita. It's
>like 102 degrees in my office. I'm outta here! :-)
>
>Priscilla
>
>
>
>
>
>
>
>
>
>>
>> By the way, TCP was developed before IP. Most people assume the
>> opposite. At the time, TCP included most of IP's current
>> functionality. Then it was wisely decided to break them up so
>> that routers could do most of IP and end hosts could do TCP. In
>> general, the divorce went well, but there were some issues that
>> never got untangled completely.
>>
>> Priscilla
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=71586&t=71556
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
