Re: UDP error handling
From: David Schwartz (davids_at_webmaster.com)
Date: 04/04/05
- Next message: Erik Max Francis: "Re: Difference between dir and ls"
- Previous message: D. Power: "make newbie, help with VPATH"
- In reply to: Barry Margolin: "Re: UDP error handling"
- Next in thread: Måns Rullgård: "Re: UDP error handling"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Mon, 4 Apr 2005 13:36:07 -0700
"Barry Margolin" <barmar@alum.mit.edu> wrote in message
news:barmar-FE27C6.15423604042005@comcast.dca.giganews.com...
>> Are you saying you have never heard of a case where a NAT box
>> 'repaired'
>> the checksum of a UDP packet that was received corrupt because it didn't
>> check the checksum before rewriting the destination address?
> No, I've never heard of a proxy modifying the payload when it doesn't
> know the application protocol.
You mean when it doesn't *think* it knows the application protocol. It
may think it knows the application protocol because of the port used (either
as source or destination in some cases) or it make 'recognize' the protocol
automatically.
>> It's hard to find URLs on the Internet because it's not clear what
>> terms
>> to search for. But I have personally dealt with many cases where proxies,
>> firewalls, and LSPs thought they understood the data I was sending and
>> made
>> manipulations that might be sensible for other protocols but made no
>> sense
>> for an arbitrary protocol layered over TCP or UDP.
>>
>> http://forums.bitpass.com/viewtopic.php?p=136
>> http://www.livejournal.com/community/lj_dev/666626.html
>> http://www.uwsg.iu.edu/hypermail/linux/net/9609.3/0024.html
>
> None of these seem to be examples of what you're describing.
These are all cases where proxies created problems because they thought
they understood what the applications wanted and were wrong. There have
definitely been reported cases of NAT boxes doing the equivalent of a 'grep'
through packets for the inside source address and helpfully changing it to
the outside address.
> I have no
> trouble believing the case where a NAT box doesn't verify the checksum
> of a received packet before doing the header rewrite.
That's really sufficient to create the problem he's worried about. A
corrupted packet could have its checksum 'repaired' by a broken NAT
application.
DS
- Next message: Erik Max Francis: "Re: Difference between dir and ls"
- Previous message: D. Power: "make newbie, help with VPATH"
- In reply to: Barry Margolin: "Re: UDP error handling"
- Next in thread: Måns Rullgård: "Re: UDP error handling"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
