| |
 |
|
|
Linux Forum Index » Linux Networking » TCP without IP...
Page 1 of 1
|
| Author |
Message |
| ... |
 Posted: Tue Aug 12, 2008 11:31 pm |
|
|
|
Guest
|
Hi,
I have read in some books as follows :-
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ? |
|
|
| Back to top |
|
| Burkhard Ott... |
| Posted: Tue Aug 12, 2008 11:34 pm |
|
|
|
Guest
|
Am Wed, 13 Aug 2008 02:31:17 -0700 schrieb aarklon:
Quote: Hi,
I have read in some books as follows :-
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ?
you would need a protocoll which can route and address packets, but an
example... |
|
|
| Back to top |
|
| goarilla at (no spam) work... |
| Posted: Wed Aug 13, 2008 4:54 am |
|
|
|
Guest
|
Burkhard Ott wrote:
Quote: Am Wed, 13 Aug 2008 02:31:17 -0700 schrieb aarklon:
Hi,
I have read in some books as follows :-
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ?
you would need a protocoll which can route and address packets, but an
example...
TCP over IPX ? |
|
|
| Back to top |
|
| Boon... |
| Posted: Wed Aug 13, 2008 5:27 am |
|
|
|
Guest
|
aarklon wrote:
Quote: Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ?
Here's a list of possible Ethernet payloads.
http://en.wikipedia.org/wiki/Ethertype
http://www.iana.org/assignments/ethernet-numbers
I don't know whether anybody uses TCP over something other than IPv4
and IPv6. |
|
|
| Back to top |
|
| David Schwartz... |
| Posted: Wed Aug 13, 2008 11:06 am |
|
|
|
Guest
|
On Aug 13, 2:31 am, aark... at (no spam) gmail.com wrote:
Quote: Hi,
I have read in some books as follows :-
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ?
It's not clear what you're asking. What are you asking?
If you're asking for examples of TCP actually being used with IP, I
doubt you'll find any. This is a theoretical possibility, but not one
that there would seem to be any reason to actually do.
The closest to an example I can think of is TCP-over-UDP
implementations that are used in special cases where TCP behavior is
wanted but TCP is not usable (because of firewalls or NAT). TCP-over-
UDP-over-STUN, for example.
DS |
|
|
| Back to top |
|
| Burkhard Ott... |
| Posted: Wed Aug 13, 2008 8:38 pm |
|
|
|
Guest
|
Am Wed, 13 Aug 2008 14:06:53 -0700 schrieb David Schwartz:
Quote: The closest to an example I can think of is TCP-over-UDP
implementations that are used in special cases where TCP behavior is
wanted but TCP is not usable (because of firewalls or NAT). TCP-over-
UDP-over-STUN, for example.
DS
But you need IP for that either and you loose the advantages of TCP. |
|
|
| Back to top |
|
| David Schwartz... |
| Posted: Wed Aug 13, 2008 10:17 pm |
|
|
|
Guest
|
On Aug 13, 11:38 pm, Burkhard Ott <n... at (no spam) derith.de> wrote:
Quote: Am Wed, 13 Aug 2008 14:06:53 -0700 schrieb David Schwartz:
The closest to an example I can think of is TCP-over-UDP
implementations that are used in special cases where TCP behavior is
wanted but TCP is not usable (because of firewalls or NAT). TCP-over-
UDP-over-STUN, for example.
But you need IP for that either and you loose the advantages of TCP.
You do need IP, but the TCP is not layered (directly) over IP. As for
loosing the advantages, no, you don't. You can keep all of the
advantages of TCP this way. (Unless you know of some advantage I'm not
thinking of.) You can still implement it in kernel space, you can
still have slow start, reordering, duplicate rejection, and so on. You
can implement it TCP-over-UDP with all of TCP's features.
DS |
|
|
| Back to top |
|
| Maxwell Lol... |
| Posted: Thu Aug 14, 2008 6:06 am |
|
|
|
Guest
|
Burkhard Ott <news at (no spam) derith.de> writes:
Quote: Am Wed, 13 Aug 2008 14:06:53 -0700 schrieb David Schwartz:
The closest to an example I can think of is TCP-over-UDP
implementations that are used in special cases where TCP behavior is
wanted but TCP is not usable (because of firewalls or NAT). TCP-over-
UDP-over-STUN, for example.
DS
But you need IP for that either and you loose the advantages of TCP.
Yes, but the OP asked about a theoretical implemention without IP.
Given that, it makes sense that either (a) there is no routing, or (b)
the user has to do their own routing using somethingother than IP. |
|
|
| Back to top |
|
| Maxwell Lol... |
| Posted: Thu Aug 14, 2008 6:11 am |
|
|
|
Guest
|
David Schwartz <davids at (no spam) webmaster.com> writes:
Quote: You do need IP, but the TCP is not layered (directly) over IP. As for
loosing the advantages, no, you don't. You can keep all of the
advantages of TCP this way. (Unless you know of some advantage I'm not
thinking of.) You can still implement it in kernel space, you can
still have slow start, reordering, duplicate rejection, and so on. You
can implement it TCP-over-UDP with all of TCP's features.
UDP uses IP for routing. So instead of
TCP -> IP
you have
TCP -> UDP -> IP
So technically it's not IP directly underneath. It's just buried a layer lower.
But it is an example of TCP over "something other than IP."
I hope the OP finds this interesting. I do. Thanks. |
|
|
| Back to top |
|
| Huibert Bol... |
| Posted: Thu Aug 14, 2008 12:15 pm |
|
|
|
Guest
|
aarklon at (no spam) gmail.com wrote:
Quote: Hi,
I have read in some books as follows :-
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ?
There is/was TCP over IPX (see rfc1791). Don't know whether it was actually
implemented.
--
Huibert
"Okay... really not something I needed to see." --Raven |
|
|
| Back to top |
|
| Moe Trin... |
| Posted: Fri Aug 15, 2008 7:43 pm |
|
|
|
Guest
|
On Thu, 14 Aug 2008, in the Usenet newsgroup comp.os.linux.networking, in
article <g81p7b$co4$1 at (no spam) registered.motzarella.org>, Huibert Bol wrote:
Quote: aarklon at (no spam) gmail.com wrote:
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ?
There is/was TCP over IPX (see rfc1791). Don't know whether it was
actually implemented.
1791 TCP And UDP Over IPX Networks With Fixed Path MTU. T. Sung.
April 1995. (Format: TXT=22347 bytes) (Status: EXPERIMENTAL)
1792 TCP/IPX Connection Mib Specification. T. Sung. April 1995.
(Format: TXT=16389 bytes) (Status: EXPERIMENTAL)
Notice the 'Status:' tag, and then look at RFC2026 section 4.2.1
4.2.1 Experimental
The "Experimental" designation typically denotes a specification that
is part of some research or development effort. Such a specification
is published for the general information of the Internet technical
community and as an archival record of the work, subject only to
editorial considerations and to verification that there has been
adequate coordination with the standards process (see below). An
Experimental specification may be the output of an organized Internet
research effort (e.g., a Research Group of the IRTF), an IETF Working
Group, or it may be an individual contribution.
Tae Sung was working for Novell in San Jose, CA (.us) and this seems
to have been a dead end - recall that Novell NetWare was an major
networking O/S, but it was pretty well locked to IPX. I know you
could get it to acknowledge the existence of IP and if you loaded the
right NLMs on the server, you could get it to _route_ IP (and Appletalk
Phase I), but I think it was some time later (after NetWare 4.1) before
they bit the bullet and actually started _running_ IP (and even when
they did, the basic directory services remained IPX).
Old guy |
|
|
| Back to top |
|
| Bill... |
| Posted: Sat Aug 16, 2008 12:14 pm |
|
|
|
Guest
|
Do you mean un numbered ip?
0.0.0.0?
<aarklon at (no spam) gmail.com> wrote in message
news:7d3e7523-a58b-4c13-8e1a-7315b7521794 at (no spam) v1g2000pra.googlegroups.com...
Quote: Hi,
I have read in some books as follows :-
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ? |
|
|
| Back to top |
|
| Joe Pfeiffer... |
| Posted: Sat Aug 16, 2008 1:34 pm |
|
|
|
Guest
|
"Bill" <bargerw at (no spam) bellsouth.net> writes:
Quote: aarklon at (no spam) gmail.com> wrote in message
news:7d3e7523-a58b-4c13-8e1a-7315b7521794 at (no spam) v1g2000pra.googlegroups.com...
Hi,
I have read in some books as follows :-
Theoretically, you could have TCP without IP, some other n/w mechanism
besides IP could deliver the data to an address, and TCP could still
verify and sequence that data
can any one give examples for this ?
Do you mean un numbered ip?
0.0.0.0?
I would assume he meant "using some other protocol completely". |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Sat Nov 22, 2008 5:59 am
|
|