On Thu, 20 Jan 2000, Mithun Bhattacharya wrote:
|Uhh isn't PMTU used for testing purposes only - I thought it wasn't
|implemented/allowed on most routers as it increases the ICMP traffic. I am sure
|some of the PPP guru's could confirm whether pppd runs with PMTU enabled. I am
|sure sending packets with the DF flag set is not a good idea for surfing the
|net, buth then it is just my opinion maybe there are reasons for running ppp
|like that.
AFAIK PMTU Discovery is very common, there are still a few sites that
don't do it, either because of a legacy TCP/IP implementation, or
because they deny ICMP packets for security reasons, real or imaginary.
Packet fragmentation is an expensive operation in terms of host loading
and PMTU Discovery helps lighten the load.
|By the way if MTU is set at 576 at your end and the otherside is trasmitting at
|1500 then someone in between will have to fragment the packet so setting the
|"Don't Fragment" flag will create problems either way.
I don't agree, here's how it looks to me:
The MTU that's set by the PPP user does not affect the PMTU for a site
sending packets to that user except to limit it for TCP/IP connections
by the TCP MSS of the PPP user.
The PPP user's MSS *is* altered to match the MTU and, for TCP/IP, a
remote site is informed of that by a SYN packet. If the site does PMTU
Discovery then it is limited by that MSS size in discovering it's PMTU.
If the site doesn't do PMTU Discovery then it should still send IP
datagrams limited by the PPP user's MSS.
Again let me emphasize that I'm not really qualified to discuss this
and anyone that can point to an error in this logic should speak up.
Actually the comp.protocols.tcpip group is where a discussion of PMTU
really belongs. There are several people there that are well qualified.
---
Clifford Kite Not a guru. (tm)
-
To unsubscribe from this list: send the line "unsubscribe linux-ppp" in
the body of a message to [EMAIL PROTECTED]
