Re: Strange MSS problem

Michael <quadfour79@yahoo.com.au> wrote:[color=blue]
> If I understand this correctly, it is not up to my gateway to send
> these packets either. I did a heap of reading on the subject and
> apparently the router that is incapable of accepting or forwarding a
> given packet (for size) should drop the packet and send the ICMP
> unreachable packet. My gateway *can't* be the cause according to
> this because it never receives these data packets back and can
> receive 1460 byte packets from elsewhere. Perhaps there is something
> I am missing?[/color]

I believe your are not understanding correctly.

Your gateway is, for all incense and porpoises the first hop IP
router. The MTU on your gateway's LAN interface is (likely) 1500
bytes. The interface on its PPP (or whatever) interface is
smaller. As such, when it receives an IP datagram with the DF bit set,
and forwarding that datagram would require fragmentation, it is
required by the IP spec to send to the src an ICMP "Datagram Too Big,
DF set" message with the size that _would_ fit without fragmentation.

It may be doing NAT, it may even be doing proxy, but it is doing all
of that transparantly, which means it is pretending to be an IP router
and must behave accordingly.

rick jones
--
No need to believe in either side, or any side. There is no cause.
There's only yourself. The belief is in your own precision. - Jobert
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

In article <Jzr5g.7016$E85.4287@news.cpqcorp.net>,
Rick Jones <rick.jones2@hp.com> wrote:
[color=blue]
> Michael <quadfour79@yahoo.com.au> wrote:[color=green]
> > If I understand this correctly, it is not up to my gateway to send
> > these packets either. I did a heap of reading on the subject and
> > apparently the router that is incapable of accepting or forwarding a
> > given packet (for size) should drop the packet and send the ICMP
> > unreachable packet. My gateway *can't* be the cause according to
> > this because it never receives these data packets back and can
> > receive 1460 byte packets from elsewhere. Perhaps there is something
> > I am missing?[/color]
>
> I believe your are not understanding correctly.
>
> Your gateway is, for all incense and porpoises the first hop IP
> router. The MTU on your gateway's LAN interface is (likely) 1500
> bytes. The interface on its PPP (or whatever) interface is
> smaller. As such, when it receives an IP datagram with the DF bit set,
> and forwarding that datagram would require fragmentation, it is
> required by the IP spec to send to the src an ICMP "Datagram Too Big,
> DF set" message with the size that _would_ fit without fragmentation.[/color]

Actually, I think he's right. Your response would be correct if he were
having problems uploading to the Internet. But IIRC his failures are
with downloads, so the router at the other end of his PPPoE link is the
one that should be sending the ICMP messages back to the remote server.

--
Barry Margolin, [email]barmar@alum.mit.edu[/email]
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***
*** PLEASE don't copy me on replies, I'll read them in the group ***

Barry Margolin <barmar@alum.mit.edu> wrote:[color=blue]
> Actually, I think he's right. Your response would be correct if he
> were having problems uploading to the Internet. But IIRC his
> failures are with downloads, so the router at the other end of his
> PPPoE link is the one that should be sending the ICMP messages back
> to the remote server.[/color]

Looking back, I believe you are right.

<Insert "Emily Litella" response here>

rick jones

I can only imagine the fun of finding where along the path between the
router at the other end of his link, and the traffic source(s) the
ICMP's, if they are indeed being sent, are being "filtered."

--
portable adj, code that compiles under more than one compiler
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

On Tue, 02 May 2006 00:48:27 +0000, Rick Jones wrote:
[color=blue]
> Barry Margolin <barmar@alum.mit.edu> wrote:[color=green]
>> Actually, I think he's right. Your response would be correct if he
>> were having problems uploading to the Internet. But IIRC his
>> failures are with downloads, so the router at the other end of his
>> PPPoE link is the one that should be sending the ICMP messages back
>> to the remote server.[/color]
>
> Looking back, I believe you are right.
>
> <Insert "Emily Litella" response here>
>
> rick jones
>
> I can only imagine the fun of finding where along the path between the
> router at the other end of his link, and the traffic source(s) the
> ICMP's, if they are indeed being sent, are being "filtered."[/color]

Finally my provider (& coworkers) admitted there is a problem and they're
investigating... You're all right on the money with the PMTu thing. Cheers :)