Hi,
True, in general I am with you here. And I agree that the base format
should be as compact and simple as possible. I am OK keeping it as is,
as long as we can make it optional for hosts as below.
Pascal Thubert (pthubert) wrote:
Hi Zach:
The 2 bits were integrated in HC when we passed the HC to 2 bytes (that
was hc_00). With hc_01, we are back to one byte and not too ready to
bounce again. I mean, there are many things we could do with more bits
in HC but then we're loosing the value of compression in the first
place.
I also don't want 2 bytes again. I understood that the dispatch space
could be used for that, but I also think we should be careful using too
much of that.
Since most times the default context is the right one, we decided to
handle the exception by putting the context information in an optional
byte that should "almost" never by there if the context is used for
renumbering only.
At that point we have that additional byte, we can shoot for 16 contexts
both directions. This might not be the best use of that byte, and we
should discuss that furthermore. But a context might not be limited to a
prefix. We can use a context to (state)fully compress an address like
the address of an edge router. One byte against 8 or 16 per packet might
be worth an entry in memory.
I have no problem with the use of the 2nd byte in the current HC-01.
Now you made it very clear that not all nodes can support 16 many
contexts. Fine. We need to work on how the option is really optional,
and how a node may not have to support all contexts. OTOH, the consensus
on 2 contexts had to do with the cost in bits in every packet. Now we
moved to a cost that is only paid by exceptional packets, 16 MAX might
become a good value after all.
We just need to find a way to make it optional as you pointed out. If we
can do that, then I am happy. 16 is a fine number, as long as a simple
host can get by with just understanding and using the default. It might
be as simple as this:
"6LoWPAN Hosts may choose to make use of the optional 4-bit context IDs
. A Host must support the default context ID. 6LoWPAN Routers must
support 4-bit context IDs."
As a simple Host will only receive incoming packets to it using the
stateless options or global default prefix. A simple host can also
choose to use only the default context when sending packets. Therefore
it doesn't have to care about the optional 4-bit context IDs at all.
I think that all this requires more thoughts. Once we really understand
what we do with contexts and what the cost is per type of context entry,
then we can figure out what the right value is, and whether all entries
are globally significant or not.
What do you think?
Pascal
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
Behalf Of Zach Shelby
Sent: lundi 13 octobre 2008 11:14
To: Julien Abeille (jabeille)
Cc: 6lowpan
Subject: Re: [6lowpan] Fwd: New Version
Notificationfordraft-ietf-6lowpan-hc-01
Hi,
I don't think the consensus changed from that 2 contexts is enough.. it
was just mentioned that with more someone may figure out what to do
with
them (renumbering). If renumbering is the only case for having a 4-bit
context identifier, then I recommend Jonathan tries to integrate a
short
1-2 bit context identifier into the base format as he mentioned is
possible, thus avoiding the 1 byte extension for 4-bit identifiers. 2-4
contexts will be plenty, and thus we wouldn't be using memory on all
nodes, or having to deal with some nodes supporting 16 and some
supporting 2. Considering that we have only a link-local context in
RFC4944, having 2-4 global contexts is already a huge improvement!
So I would give like to see 1-2 context ID bits for source/dest in the
base header, as a priority over supporting a large range of multicast
address types.
- Zach
Julien Abeille (jabeille) wrote:
Hi Zach,
I second your concern
There was recently a discussion specifically about the number of
contexts needed.
There seemed to be a consensus that multiple contexts was useful for
renumbering only, which is rare, and that 2 contexts was enough.
Was there new input since then which justifies the use of 4-bit
contexts?
Cheers,
Julien
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
Behalf Of Zach Shelby
Sent: vendredi 10 octobre 2008 14:58
To: Jonathan Hui
Cc: 6lowpan
Subject: Re: [6lowpan] Fwd: New Version Notification
fordraft-ietf-6lowpan-hc-01
Hi,
Here are my comments on hc-01. Overall I think we are quite close
with
the comments from Pascal and Julien.
I found the DDF name confusing. Now it has been renamed CTX in the
latest formats on the list, great.
As Pascal pointed out, nodes need to be able to decode all these
options. One thing that worries me in having 16 contexts, is that
nodes
are forced to store them all. That is worst case 272 bytes of memory
consumption. Worth it? Is it possible that an ultra-simple node would
only have to understand the default context?
The default context identifier '0000' could be clarified more in the
draft. When CTX is '10' then the default context identifier '0000' is
used. It should be made clear that this is the prefix advertised in
RAs
as useable for SAA.
Regarding the ND for 6lowpan draft, HC-01 means we need to add a
4-bit
context identifier flag to the 6LoWPAN Prefix Information Option. To
make things clear that should always be '0000' if the A flag is '1'?
- Zach
Jonathan Hui wrote:
http://tools.ietf.org/html/draft-ietf-6lowpan-hc
This new version reflects the recent discussions on the list. The
changes are:
- HC back to 1 byte by default by stealing a few bits from the
dispatch field.
- Added better support for multicast address compression.
- Fixed alignment for UDP port compression.
- Better support for Traffic Class and Flow Label compression.
- Pascal joined as an author.
Happy reading!
--
Jonathan Hui
Begin forwarded message:
From: IETF I-D Submission Tool <[EMAIL PROTECTED]>
Date: October 8, 2008 7:15:35 PM PDT
To: [EMAIL PROTECTED]
Cc: [EMAIL PROTECTED]
Subject: New Version Notification for draft-ietf-6lowpan-hc-01
A new version of I-D, draft-ietf-6lowpan-hc-01.txt has been
successfuly submitted by Jonathan Hui and posted to the IETF
repository.
Filename: draft-ietf-6lowpan-hc
Revision: 01
Title: Compression Format for IPv6 Datagrams in 6LoWPAN
Networks
Creation_date: 2008-10-09
WG ID: 6lowpan
Number_of_pages: 15
Abstract:
This document specifies an IPv6 header compression format for IPv6
packet delivery in 6LoWPAN networks. The compression format relies
on shared context to allow compression of arbitrary prefixes. This
document specifies compression of multicast addresses and a
framework
for compressing next headers. This framework specifies UDP
compression and is prepared for additional transports.
The IETF Secretariat.
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan
--
Zach Shelby
Head of Research
Sensinode Ltd.
Kidekuja 2
88610 Vuokatti
FINLAND
mobile: +358 40 7796297
This e-mail and all attached material are confidential and may contain
legally privileged information. If you are not the intended recipient,
please contact the sender and delete the e-mail from your system
without
producing, distributing or retaining copies thereof.
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan
--
Zach Shelby
Head of Research
Sensinode Ltd.
Kidekuja 2
88610 Vuokatti
FINLAND
mobile: +358 40 7796297
This e-mail and all attached material are confidential and may contain
legally privileged information. If you are not the intended recipient,
please contact the sender and delete the e-mail from your system without
producing, distributing or retaining copies thereof.
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan
