Re: shim header in every shimmed packet

[Iljitsch van Beijnum <iljitsch@muada.com>]

On 12-okt-2005, at 11:00, marcelo bagnulo braun wrote:

> > Considering all of this, I think adding a bit and a single insert  
> > shim header / don't insert shim header decision into the core spec  
> > and therefore (presumably) all implementations isn't too much to ask.

> Let me see if i understand:
> Do you want that the base spec explicitly adds a bit in the context  
> establishment exchange that means that insert/don't insert shim  
> header and nothing else so that in other document where the  
> mechanisms for suppressing the shim header can use that?,

This sentence isn't all that clear to me...

> or
> Do you want to include both the bit in the establishment exchange  
> and the mechanisms used to suppress the shim header in the core  
> shim spec?

> I mean are we talking about adding a single bit for future  
> extensibility or the bit plus the complete sets of mechanisms in  
> the core spec?

What I want is:

- one bit (or whatever makes sense after we look at the details) that  
a host can use to tell its correspondent that it doesn't want to see  
a shim header in incoming packets for the context in question
- when indicated by the bit and within some constraints (for  
instance, there are no intermediate headers between the IPv6 header  
and the ULP), hosts implementing the spec are REQUIRED to leave out  
the shim header for packets belonging to this context
- advice that hosts SHOULD set the bit such that the shim header is sent

Under which circumstances the shim header can safely be left out so  
hosts can set/clear the bit can then be explored in a separate document.