Re: CVS commit: src/sys/netinet

[Daniel Carosone <dan%geek.com.au@localhost>]

On Mon, Sep 08, 2003 at 07:47:51PM +0100, David Laight wrote:
> > the net.  You also don't need to consume IPid space -- or at least,
> > you don't need to worry about preventing duplicates -- on packets
> > that have  set.
>             ^ dont fragment ???

Yes, I guessed that's what he meant, too.

> Actually there is is little point separating on the src address [1]

Perhaps more so in v6 land, but still generally a good point.

> I also suspect that very few IP packets are fragmented anywhere other
> than the source system (pppoe is one place where it could be done...) [2].

If PMTU is on and working, sure.  Otherwise, there are plenty of
places where things can get fragmented. Some of these are becoming
more common, not less.

The number and types of media, and their corresponding frame size
properties, that can and do carry IP traffic, are growing. At least
the following can carry larger frames than ethernet, and so will
be an issue if the source is on one of these:
 - gig-e
 - fw-ip
 - fc-ip
 - hippi
 - (as you point out) fddi or Token Ring (and yes, some people
   still are using it and even deploying more.

And then there's the number of things that are getting tunnelled
inside other things, so even if everything had ethernet MTU at
layer 2, there's encapsulation overheads across bits of WAN.

 - pppoe tunnels
 - ppp serial links
 - IPsec
 - MPLS
 - GRE
 - L2TP
 - Mobile-IP/STF/etc..

> So you don't really need to worry too much about the IPid on unfragmented
> packets.

But how do you know which ones they are?

--
Dan.