From: Howard C. Berkowitz (hcb@gettcomm.com)
Date: Wed Apr 13 2005 - 10:57:14 GMT-3
At 12:31 PM +0200 4/13/05, Jelle Borsje wrote:
>Hej,
>
>When it comes to the configuration, and usage, it is
>almost the same. The encapsulation type on the tunnel
>interface is ipinip, instead of gre.
>
>In reality, IPinIP simply puts an IP header in front
>of the original packet. The outside IP header, has
>protocol type 4 and the source and destination
>addresses are the tunnel endpoints. Since there is no
>'customized' header, there is no field to indicate a
>protocol type, unlike GRE.
>
>The GRE header is an additional header that is placed
>between a new (outside IP header) and the payload (the
>original packet we want to travel through the tunnel,
>which could well be an IP packet). The outside IP
>header contains source and destination IP addresses of
>the tunnel endpoints. The GRE header contains a number
>of fields, interestingly a protocol type field. This
>means the payload could be many different types of
>protocols, not just IP. This means that we can tunnel
>other protocols in GRE tunnels, not only IP. IPinIP
>tunnels can only tunnel IP traffic as far as I know.
Correct. Multiprotocol use is the first intention of GRE. In
general, it was simplified over time; compare the current
http://www.isi.edu/in-notes/rfc2784.txt against the original RFC 1701
and 1702.
I happened to attend the Oslo IETF meeting where the initial,
procedural need to update the original RFC's came up. The
Unidirectional Link Routing RFCs assume IPIP or GRE, but the working
group was surprised that GRE was only an Informational RFC, and
Standards Track RFCs, such as UDLR, must reference only standards
track. When the update to 2784 was done, the team removed some
1701/1702 features that never had found practial use.
UDLR, incidentally, isn't usually truly unidirectional. Its main use
comes in satellite video transmission, where a high-bandwidth
satellite channel is used for content, and a slow terrestrial channel
for ACKs if TCP is used.
There has been continuing, if low-level, interest in using the
inherent multiplexing capability of GRE, for people who don't want to
use MPLS. To do so would require redefining the use of the Protocol
Type field as a sub-tunnel indicator (probably a bad idea) or
starting to use the Reserved1 field as an IPv4 (or IPv6) sub-tunnel
ID.
>
>I think you would have a hard time getting ISIS (since
>it uses CLNS) to work over IPinIP encapsulated
>tunnels, while it works over GRE. IPX can also be
>transported in a GRE tunnel.
>
>When it is just IP traffic you are looking at, the
>difference between them is probably not striking...
>when you want to run other protocols than IP through
>the tunnel, it becomes more appearant.
IPIP does have a little less overhead, and, for some UDLR
applications, bandwidth is a very scarce resource.
>Hope this helps a little.
>
>Greetz
>Jelle
>
>--- stephen skinner <stephenski@gmail.com> wrote:
>> hello ,
>>
>> i was recently doing the NMC self-assessment .and
>> there was one topic
>> on there that i don`t know and i have tried in vain
>> to find out some
>> info .
>>
>> what is the difference between IPinIP and GRE
>> tunnels
>>
>> TIA
>>
>> Steve
>>
>>
>_______________________________________________________________________
>> Subscription information may be found at:
>> http://www.groupstudy.com/list/CCIELab.html
>
>_______________________________________________________________________
>Subscription information may be found at:
>http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Tue May 03 2005 - 07:54:57 GMT-3