Re: DoCoMo Implementation Issues with CTP

Hi Rajeev,

Please find my comments below

regards,
Raghu


>
> Jim,
>
> the suggested changes are okay. I am curious about couple
> of suggestions however..
>
> - When both IPv4 and IPv6 addresses are used in CTD,
>  "Processing the address of the opposite
> version in the stack is complicated. " Could you elaborate on
> this ? I can see that once IP (v4 or v6) stack demultiplexes the
> packet to belong to a CTP client, it would just forward the block
> to a CTP module (a daemon for instance), which could then
> process contexts associated with both the addresses.
>
IP Version flag in the draft indicates whether the IP Version
can be either IPv4 or IPv6 or both(IPv4 & IPv6).
When MN is assigned both IPv4 & IPv6 addresses,
it is assumed that AR also has both IPv4 & IPv6 addresses.
When MN moves, sending two CT messages for the same MN
is unnecessary. Hence to avoid this, either IP version flag
should not support both IPv4 & IPv6 at the same time or
the draft should mention, which IP version to take precedence.


> - Isn't MTU a consideration in increasing the size of the
>     Length field to 16 bits ?
>

Since CTP is an application layer program, I guess
supporting upto 16bits is more reasonable for the
following reasons
1. UDP supports upto 16bits
2. As more Feature profiles are defined more
    contexts in the same CTP needs to be transferred.

As per the sections 2.4 & 2.5 in the draft
"message length in units of 8", Padding bytes
 are required for both CDB & CT Messages.
These padding bytes get critical when we are
reaching the maximum size ie 2048 bytes.




> Thanks,
>
> -Rajeev
>
>
> James Kempf wrote:
>
> > DoCoMo is in the process of implementing CTP, and we came up with the
> > following list of issues in the spec. Since none of them seem
particularly
> > large, Allison suggested that we simply update the current version of
the
> > document (which is now under review by the ROHC WG) in place, without
reving
> > the document to a new version. She will make the changes.
> > Below, a list of the issues and suggested changes to the document to
address
> > them.
> >
> > Any comments? Please let me know by March 6.
> >
> >             jak
> >
> > -------------------------------------------
> >
> > 1) Section 2.5.1, pg. 10; Section 2.5.2, pg. 11; Section 2.5.3, pg. 12;
> > Section 2.5.4, pg. 13; Section 2.5.6, pg. 15 - 'V' flag
> >
> > The spec allows CTP messages to carry both IPv4 and IPv6 addresses, but
as a
> > practical matter in the implementation, a CTP message will be sent to
the
> > IPv4 or IPv6 address of a router. Processing the address of the opposite
> > version in the stack is complicated. Suggestion is to only allow IPv4 or
> > IPv6 addresses, but keep the flag field size at 2 bits in case further
> > experience shows that there are reasons why both might be sent. The
router
> > can issue separate messages if both addresses must be changed,
Suggested
> > change in text:
> >
> > Replace: "When set to '10', indicate presence of both IPv6 and IPv4
Previous
> > (New) addresses."
> > With: "Values of '10' and '11' are reserved".
> >
> > 2) Section 2.5.3, pg. 12 - Length of Algorithm field in CTD message
block.
> >
> > The length of the algorithm field in the CTD message block is shown as
17
> > bits. It should be 16.
> >
> > 3) Section 2.5.1, pg. 10; Section 2.5.2, pg. 11; Section 2.5.3, pg. 12;
> > Section 2.5.4, pg. 13; Section 2.5.6, pg. 15 - Length field.
> >
> > The size of the message length field is currently 8 bits, and the units
for
> > the length are in 8 octet words. This allows a maximum of 2048 octets,
which
> > might not be enough for bundling large amounts of context if padding is
> > required. Suggestion is to increase the size of the field and make it be
> > units of octets, for a total of 65535 octets. This would only involve
moving
> > the flags into the Reserved field, and would not otherwise modify the
> > message.
> >
> > Replace: In message diagram blocks, move Length to be 16 bits rather
than 8.
> >
> > Replace: "Length    message length in units of 8 octet words"
> > With: "Length    message length in units of octets"
> >
> > 4) Section 2.5.1. pg. 9; Section 2.5.2, pg. 11; Section 2.5.3, pg. 12;
> > Section 2.5.4, pg. 13; Section 2.5.5, pg. 14; Section 2.4.6, pg. 15 - IP
> > address field size.
> >
> > The message block diagrams show the size of the IP address fields as 4
> > octets, but if the IP address is for IPv6, it will be 16.
> >
> > Replace: In the message block diagrams, replace the final "|" with a "~"
to
> > indicate a variable length field. Indicate variable length field by
putting
> > "(4 or 16 octets)" after the text in the field.
> >
> > 5) Section 2.5.3, pg. 12 - Key field in message diagram
> >
> > Only 4 bytes are shown for the key field, but the key size field allows
> > 65535. The message diagram needs to be changed.
> >
> > Replace: In the message diagram, replace the final "|" with a "~" to
> > indicate a variable length field. Indicate the variable length field by
> > putting "(variable)" after the text in the field.
> >
> > _______________________________________________
> > Seamoby mailing list
> > [email protected]
> > https://www1.ietf.org/mailman/listinfo/seamoby
>
>
> _______________________________________________
> Seamoby mailing list
> [email protected]
> https://www1.ietf.org/mailman/listinfo/seamoby
>