Pantek Library
Hosting Provided By
CybrHost
High Speed Hosting

Re: [dtn-interest] Demux tokens (was Re: Comments on DTNRG documents from BBN SPINDLE project)

From: Scott Burleigh <Scott.Burleigh(at)jpl.nasa.gov>
Date: Mon Apr 11 2005 - 20:15:42 EDT

Michael Demmer wrote:
>>I see your point about the dictionary, so I would be willing to go

> 

>>For the Demux header, though, the whole point of the header is bit
> 
> Ok...
> 

>>>There's the fragmentation header, which based on the other discussion
> 
> Why? Can't the receiving bundle agent simply forward the fragment no
> differently than if it were a full bundle?

Maybe I misunderstand exactly what we mean by the term "optional". Certainly a bundle agent that is only forwarding a bundle doesn't need to do fragmentation or even recognize that a given bundle is a fragment. But any given bundle agent is potentially the final destination agent for a given bundle, and if the final destination agent doesn't recognize fragments and reassemble original payloads from them then I would say you've got a failure of interoperability. That is, we can certainly say that fragmentation is optional, but I would suggest that reassembly isn't.   (Nor is the -- what shall we call it? -- responsive fragmentation that a custodial agent must do when it receives a custody acceptance signal for a fragment of an original bundle for which it is the current custodian.)

>>And especially since this is a fixed-size header, I think I'd find

> 
> Yeah -- I would agree. The only argument here is whether support for
> fragmentation in any way (including parsing the header) is a
> requirement in the basic protocol spec.

Unfortunately I think it has to be.

>>I think the impulse here is right, but the Demux and Fragment

> 
> Again, I think they are actually optional. If I'm convinced otherwise,
> then I would agree with you that they should be exceptions and have no
> need for the explicit length preamble, as they are actually fixed
> length.

Well, the proposed Demux header wouldn't be fixed-length (those SDNVs), but I think we can agree on that structure anyway: I'm fine with having an 8-bit header length in the header.

Scott



dtn-interest mailing list
dtn-interest@mailman.dtnrg.org
http://mailman.dtnrg.org/mailman/listinfo/dtn-interest Received on Mon Apr 11 20:21:43 2005
Do you need help?X

This archive was generated by hypermail 2.1.8 : Wed Aug 23 2006 - 13:27:02 EDT


Contact Us  Legal Notices  Order Services Online 
Pantek Home  Privacy Policy  IT news  Site Map  Pantek Library