[olug] bad practices at home - not relevant!

David Walker linux_user at grax.com
Sun May 27 15:42:22 UTC 2001


Using the chart that shows X25 MTU = 576

Outgoing:
wouldn't the ICMP from the X25 network pass through the router with the
private IP address telling the originating @home machine to use an MTU of
576?

Incoming:
the packet hits the X25 network before it ever gets to anywhere on the @home
network so it already uses an MTU of 576?

-----Original Message-----
From: puzzled [mailto:puzzled]On Behalf Of puzzled
Sent: Sunday, May 27, 2001 9:08 AM
To: olug at bstc.net
Subject: Re: [olug] bad practices at home - not relevant!




     @home user -> internet -> iraqi FTP server on their old X25 network


    or a more cheese filled example

   @home user -> internet -> ISP run by pissed off @home customer who policy
routes their customers over a 128 byte MTU link


   :-)  Not that either of these exists in more than one place on the
internet
today, but it could spread :-)




David Walker wrote:

> If the MTU values given are fairly standard then outgoing packets should
> never have a problem since they will always be ethernet packets and are
not
> likely to be routed over an X.25 network, therefore they will not ever hit
a
> router that cannot handle the MTU of 1500 since that is the lowest likely
> MTU.
>
> Incoming packets must hit an MTU equal to or less than the router with the
> internal IP prior to hitting that router so that that particular router
will
> never have to fragment their packets or reject them.  @home can count on
the
> final destination of virtually all packets on their network being ethernet
> so couldn't they adjust their routers accordingly?
>
> -----Original Message-----
> From: puzzled [mailto:puzzled]On Behalf Of puzzled
> Sent: Saturday, May 26, 2001 11:27 PM
> To: olug at bstc.net
> Subject: Re: [olug] bad practices at home - not relevant!
>
>      :-)
>
>    A much deserved poke - I've been asked before if the use of RFC1918
> (aka private) addresses was an issue in large scale ISP operations and
> I've always said that it didn't matter - I'd never really considered the
> path MTU discovery aspect.
>
>    I've known the mechanics of PMTU for a long time but I've never seen
> an instance of it in action and I spend *a lot* of time watching traffic
> with various protocol analyzers.
>
>    Thus far this has been a pretty theoretical discussion - anyone got a
> real world example they can share? I'd be happy if someone could come up
> with a client/server interaction that shows actual PMTU discovery. If
> this can be found I'll set up a unix box on a path where we can vary the
> MTU and we'll do some benchmarks on network performance.
>
> Brian Roberson wrote:
>
> > Just an FYI on my quick response & flame on the subject:
> > Neal & I ( puzzled at home.com ) go way back and this was a poke at him,
> > not at the group.
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: olug-unsubscribe at bstc.net
> > For additional commands, e-mail: olug-help at bstc.net
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: olug-unsubscribe at bstc.net
> For additional commands, e-mail: olug-help at bstc.net
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: olug-unsubscribe at bstc.net
> For additional commands, e-mail: olug-help at bstc.net


---------------------------------------------------------------------
To unsubscribe, e-mail: olug-unsubscribe at bstc.net
For additional commands, e-mail: olug-help at bstc.net



---------------------------------------------------------------------
To unsubscribe, e-mail: olug-unsubscribe at bstc.net
For additional commands, e-mail: olug-help at bstc.net



More information about the OLUG mailing list