TCP/IP vs OSI?

Hi Can I say the following

1) TCP/IP layers model is similar to OSI layers model. I mean TCP/IP
is like an implementation to the OSI model which is in its turn, a
general model to make a network? So to make a network you need
software/hardware that are together work as implementation of TCP/IP

2) And Operating Systems which are used TCP/IP (as part of OS) can be
called Netwroking OS or Operating System that can do networking??

Am I right?? thanks a lot.

"esara" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed) m...
> Hi Can I say the following
>

Its really off topic. Try the comp.protocols hieeracy

> 1) TCP/IP layers model is similar to OSI layers model.

Similar.
>I mean TCP/IP
> is like an implementation to the OSI model which is in its turn, a
> general model to make a network?

First remember TCP/IP pre-dates OSI implementations . So as it existed
before OSI it can't be an implementation of OSI.

In fact its really not "like" an implementation at all. When you are running
TCP/IP on Ethernet the layers don't even match properly. You have to fudge
the bottom and shoe-horn the MAC and the LLC into one layer. Complete yuk.
And in TCP/IP there is no concept of session or presentation.

So to sum up, TCP/IP, or more correcty the Internet Protocol Suite is a set
of layered protocols. It has some simalarities to the OSI but there is no
exact corrospondance between the number and function of the layers.

> So to make a network you needj
> software/hardware that are together work as implementation of TCP/IP
>

NO, you need software and hardware. Does not have to be TCP/IP. It could
equally be NETBIOS, IPX/SPX, X.25, SNA or OSI.

> 2) And Operating Systems which are used TCP/IP (as part of OS) can be
> called Netwroking OS or Operating System that can do networking??
>

I think the term "Networking OS" is pretty redundant these days. It used to
be applied to things that were servers like LAN Manager and Netware (both
of which can be used without TCP/IP), LINUX or Solaris. ( which I think both
need TCP/IP )

> Am I right?? thanks a lot.

Mostly wrong Sorry.

Dave

P.S. I usually charge £10 a question for doing homework :-)

"<Snip>
>
> First remember TCP/IP pre-dates OSI implementations . So as it existed
> before OSI it can't be an implementation of OSI.
<SNIP>

I seem to remember this being the other way around: the 5 layer OSI model
came about pre-1984 and was used as the basis for may comms systems
including X.25, which mapped onto the transport layer and X400 which went
one layer up above that. The IEEE 802 project came along and then
refined/re-invented the lower two layers to map on DLC, MAC and Physical
layers.

Mike

"fred bloggs" <(E-Mail Removed)> wrote in message
newsbLrc.6025$(E-Mail Removed)...
>

Mike,
I know I should rise to the bait but :-)

> "<Snip>
> >
> > First remember TCP/IP pre-dates OSI implementations . So as it existed
> > before OSI it can't be an implementation of OSI.
> <SNIP>
>
> I seem to remember this being the other way around: the 5 layer OSI model
> came about pre-1984 and was used as the basis for may comms systems

I don't beleive that people used the model as basis for comms systems. The
the first draft of OSI that I can find reference to is dated 1977. Most of
the networks usually quoted as being examples of the OSI model were already
mature at least in design if not in implemenation by this time.

The two biggies were "ARPA Net" which is where the internet started. By this
time it was already seven years old. RFC1000 makes it clear that ARPAnet was
developed using layers without reference to OSI. It says:-

"If we had only consulted the ancient mystics, we would have seen
immediately that seven layers were required."

The other network which used to often be compared to OSI was SNA. Again this
was announced in 1974 as

http://www-1.ibm.com/ibm/history/doc.../1970-1984.pdf

makes clear. Again well before OSI was even in draft.

From what I saw and expereinced at the time the reverse happened. Because
OSI was predicted to become the dominant communications model, and because
its provision (at least in Europe) was expected to become an essentioal part
of computer and network procuremnet, folks went round in headless chicken
mode trying to show that their network conformed to the OSI model, and that
they shouldn't worry. So IBM produced things showing how SNA used the same
layers as OSI (it doesn't really) and folks attempted to fit TCP/IP onto
same model.

At the lower layers the comparisions usually work quite well. All networks
have some kind of standard physical interface, be it a electrical or
optical. To transfer data you need to send it down a link, so you need a
link level protocol, then if you have multiple links you need network
protocol to manage end-to-end data flow. So the lower layes usually fit
well.

However you may not need a transport layer. There are a whole host of
applications which sit directly on top of X.25 and work happily with no
transport layer. I used X.29 terminals for many years with no transport,
session or presentation layer. In fact for most purposes session and
presentation are superfluos. We don't use them on the Internet, and in
particular HTTP/web was designed to be "session free".
..

> including X.25, which mapped onto the transport layer and X400 which went
> one layer up above that.

X.25 also pre-dates OSI. It maps onto the network layer, not transport
layer. However because X.25 provides reliable packet delivery it is normally
used with a NULL transport layer (Transport Class 0). X.400 is an
application so should map onto layer 7. It too was standardised before OSI
was properly complete so it originally sat directly on top of X.25, and can
be used over null session and presentation control layers.

> The IEEE 802 project came along and then
> refined/re-invented the lower two layers to map on DLC, MAC and Physical
> layers.
>

As I said they fudged layer two because DLC and MAC are squashed into one
layer.

> Mike
>
>

Any with OSI dead I don't see why we still try and shoe horn network stacks
into the OSI model. Different networks need different sets of protocols. As
we have seen from the 802 project we may need additional layer on WANs over
LANs. Circuit Switched (x.21) networks don't need separate link and network
layers, as the network provides an end-to-end link for each connection. As
an example its o.k. As model into which all networks fit, it was obsolete
before it was produced.

Dave

> Mike,
> I know I should rise to the bait but :-)
>
Hi Dave

No bait intended - that's not my 'style'.

I agree with you that everyone looked at the OSI stack and thought - "Hm
that looks good, lets crowbar our stuff into it". There were a couple of
things in your reply that I didn't know/appreciate so thanks for that.

Kind regards, Mike