I guess that I see your point but if we are not going to talk about
implementation details, why do we do so at the end of section 5.3. It
seems that much of that is implementation dependent.
Also is the Note: in section 5 right before 5.1 actually necessary.
geoff
On Thu, 2007-01-18 at 01:00 -0800, gabriel montenegro wrote:
> TTL (rather "Hop Count") of 1? Sure. The IPv6 layer handles that, not
> lowpan.
> Intermediate? No. ICMP is an IPv6 function. Intermediate nodes (as in
> a lowpan mesh)
> must not interfere with this. Now, if we wanted to define a lowpan
> layer error
> notification mechanism (analogous to ICMP at the IP layer), we could.
> But that's
> not ICMP, and I'd argue that we should define that in a separate draft
> later on
> (perhaps as part of forthcoming mesh routing specifications).
>
> Sure, layering may be relaxed for optimization reasons. But this is
> nothing
> new: it's done today in optimized implementation regardless of lowpan.
> Still,
> the conceptual model holds, so those are implementation details we
> shouldn't
> be concerned about in this particular document.
>
> -gabriel
>
> ----- Original Message ----
> From: Geoff Mulligan <[EMAIL PROTECTED]>
> To: gabriel montenegro <[EMAIL PROTECTED]>
> Cc: 6lowpan <[email protected]>
> Sent: Wednesday, January 17, 2007 11:57:00 PM
> Subject: Re: [6lowpan] Slight problem with format document
>
> Gabriel,
> If a node within a 6lowpan network receives a packet with a ttl of 1
> won't it send back an icmp error message if it was supposed to forward
> the packet, either if you were routing between 6lowpan nodes or
> between
> 6lowpan networks. Also couldn't an intermediate node send back a
> "parameter problem" icmp message.
>
> And since these are being implemented on micro-controllers is it
> strictly true that the layer separation is completely observed. I
> could
> see someone implementing a stack where you would not waste the limited
> ram space to uncompress the ip header and that the ip layer would deal
> with and understand the compressed headers and in this case the
> transport header then might still be compressed.
>
> Additionally I think that it is necessary to clarify if the checksum
> is
> run over the compressed or uncompressed ip pseudo header.
>
> I think for sake of clarity one short section (I may have been wordy)
> isn't bad, unless it is just wrong.
>
> geoff
>
>
>
> On Wed, 2007-01-17 at 23:07 -0800, gabriel montenegro wrote:
> > Is this even necessary to be said? ICMP msgs will never be sent by
> > intermediate nodes
> > in a lowpan mesh, but only by the IPv6 end nodes as part of normal
> > IPv6 processing.
> > This processing takes place after the lowpan layer delivers packets
> to
> > the
> > IPv6 stack. So by the time the IPv6 stack gets the packet, it will
> > have been
> > decompressed. This is the same for lowpan header compression or any
> > other type
> > of header compression.
> >
> > So I wonder if the text below is needed. It would be needed if the
> > entity sending
> > back ICMP messages was the lowpan layer. Yes, it would see
> compressed
> > headers.
> > But the entity is the IPv6 stack, which does not see those
> compressed
> > headers.
> >
> > Comments? Am I missing something?
> >
> > -gabriel
> >
> > ----- Original Message ----
> > From: Geoff Mulligan <[EMAIL PROTECTED]>
> > To: 6lowpan <[email protected]>
> > Sent: Wednesday, January 17, 2007 10:36:52 PM
> > Subject: [6lowpan] Slight problem with format document
> >
> > While reviewing the format document I realized that we didn't
> describe
> > handling of ICMPv6 error messages. Since the 6lowpan headers may be
> > compressed it is necessary to uncompress the original IP and
> transport
> > headers before sending the ICMP error message.
> >
> > Here is my proposed text for a new section 12...
> >
> > 12. Handling ICMPv6 messages
> >
> > ICMPv6 (RFC2463) is used to report errors and carry IP layer
> > information
> > and functions such as diagnostics. There are two groups of ICMPv6
> > messages: error messages and informational messages. Each message
> > consists of a type field (if the high order bit is 0 it is an error
> > message), a code and checksum field. These fields are followed by
> the
> > ICMPv6 message body. For ICMPv6 error messages (Type <128) the
> > message
> > body shall contain as much of the original (offending) IP message
> > without exceeded the minimum IPv6 MTU.
> >
> > As described in the preceding section (Header Compression), the
> > original
> > IPv6 and Transport (TCP or UDP) headers may be compressed via HC1
> and
> > HC2 encoding. So that the destination node of an ICMPv6 error
> message
> > can properly process the message the source node must decompress the
> > IPv6 and transport headers before sending the ICMPv6 message. This
> is
> > necessary because the original MAC addresses and the HC1 and HC2
> > encoding headers will be lost and the recipient would have no
> ability
> > to
> > reconstruct the original message nor compute the proper checksum.
> >
> > The ICMPv6 message itself is carried inside and IPv6 packet and this
> > packet may be transmitted over the 6LoWPAN network utilizing header
> > compression. ICMPv6 messages that are larger than the available
> > payload
> > of the 6LoWPAN network will need to be fragmented. Assuming the
> > maximum
> > frame overhead, maximum link layer security and including a 6LoWPAN
> > Mesh
> > Header, Fragmentation Header and Dispatch Header, sending the
> > uncompressed IPv6 and transport headers should still allow for 25
> > octets
> > of the original packet payload. Packets using short addresses, no
> > security, and just a 6LoWPAN Dispatch header will be able to carry
> 61
> > octets of the original packet.
> >
> >
> >
> >
> > _______________________________________________
> > 6lowpan mailing list
> > [email protected]
> > https://www1.ietf.org/mailman/listinfo/6lowpan
> >
> >
> >
>
>
>
>
_______________________________________________
6lowpan mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/6lowpan
