Maximum Transfer Units (MTU) have an significant impact on the efficiency of traffic
flow. MTU's are set on a per link basis and describe the maximum datagram size
permitted on a link. Should a datagram size exceed the particular MTU on a link, the
datagram is either dropped or fragmented depending on the state of the DF (do not
fragment) bit in the datagram header. In the even of a drop, an ICMP Destination
Unreachable message is sent from the router who dropped the datagram to the source.
MTU path discovery involves a process where the source tries to figure out what the
lowest MTU is across a set of links from source to destination. Figuring this out
allows the pending transmission to be optimized from an MTU perspective. The process
as described in RFC 1191, indicates that a source will send a datagram with the DF bit
set (ie do not fragment) and an MTU equal to the size of its Next_Hop router which it
already knows. Should this MTU be the lowest, the transmission will succeed. Should
another MTU be lower along the path, an ICMP message indicating a need to unset the DF
bit will be returned by the particular router with the lower MTU setting. Upon
receiving this message, the source can either retest with a lower MTU, or decide to
unset the DF bit.
However, should that particular router happen to have a link address out of the 1918
block, the likelihood of the source ever receiving the ICMP notification is
significantly diminished due to best practises filtering policies which hopefully have
been enacted with other AS's. Hence, the source will be unable to successfully
complete this process.
Hope that helps
Pete
*********** REPLY SEPARATOR ***********
On 2/26/2001 at 10:44 AM Kane, Christopher A. wrote:
>As part of this thread, several people have mentioned that one of the
>problems created is "breaking MTU path discovery." Could someone explain
>what this means?
>
>Thanks
>
>-----Original Message-----
>From: Howard C. Berkowitz [mailto:[EMAIL PROTECTED]]
>Sent: Sunday, February 25, 2001 11:21 PM
>To: [EMAIL PROTECTED]
>Subject: Re: Private Internet Addressing
>
>
>>
>>and the reason an ISP would be considered "clueless" for using RFC1918 on
>>internal point to points is..........?
>>
>>Brian
>>
>
>Let's see...
>
>It confuses troubleshooting because valid routes may appear to be
>looping, with the same address traversed more than once.
>
>The addresses can't be resolved with reverse DNS.
>
>It breaks MTU path discovery.
>
>It violates the spirit of RFC 2827 and reverse path verification.
>
>_________________________________
>FAQ, list archives, and subscription info:
>http://www.groupstudy.com/list/cisco.html
>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
>
>_________________________________
>FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
>Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
