[hpsdr] Packet Length and Fragmentation

[Jeremy McDermond]

Sorry I haven't been around the TS sessions for the last couple of weeks, but I've been at various ham conferences over the last couple of weeks, and will be at DCC next weekend.

I was listening to the Teamspeak session from last Friday and noticing Dave WA7YWQ's comments, and have some contributions.

If you're getting fragmentation on the same network segment, you have some issues going on.  Yes, any router in the chain is allowed to change the packet length.  In IP parlance, we call it "fragmentation," and it's something that we try to avoid like the plague in network designs because it tends to destroy routers if there's a lot of fragmentation.  This is because the router has to handle the fragmentation and that isn't usually on the fast forwarding path.

The only reason why you fragment packets is because the MTU of two interfaces on a router are different.  For example, if you have ethernet on one end and a OC3 on the other, where the MTU of Ethernet is 1500 bytes, and the MTU of the OC3 is 4470 bytes.  If the router receives a packet on the OC3 that is larger than the 1500 byte Ethernet MTU, the router will have to fragment the packet so that it can be transmitted on the Ethernet.  This rarely happens, because both endpoints are usually on 1500 byte MTU ethernet, and IP doesn't "coalesce" packets.  Once they're fragmented, they stay fragmented.

So, the upshot of this explanation is that if you're on the same wire, there is no possibility for fragmentation, and you should receive packets exactly as they're transmitted.  The only time when you'll get modified packets is when you go through a router.  Switches are a L2 device, and will never fragment.

--
Jeremy McDermond (NH6Z)
Xenotropic Systems
mcdermj at xenotropic.com