Hello Stig,
Bear with me for the delayed reply. Do look inline for my reply to your
comments and suggestions.
On Thu, May 12, 2005 at 02:46:10PM -0700, Arun Thulasi wrote:
Hello All,
This is a request to review and comment on the internet draft available
at
http://www.ietf.org/internet-drafts/draft-arunt-ipv6-multicast-filtering-rules-00.txt
The draft is written to explain a set of behaviors in IPv6 multicasting
scenarios. Since the behavior seemed to vary on different
implementations, this draft specifies both the behaviors and puts forth
one of them as preferred.
Do look into the draft and offer your comments.
The behaviour described in 5.2 pretty much matches what I would expect.
However, you don't say anything about SSM or the RFC 3678 source filering
API.
I did a run through of the ssm and allied drafts. The behavior that we
have suggested in the draft is independent of SSM. By this I suggest
that if the user wants to implement SSM, it would have to involve some
kind of filtering. Since SSM requires a (S,G) format where there has to
be exactly one source and multiple receivers (as specified in the
document), it is different from the default behavior what we propose in
our draft which applies for ordinary multicasting scenarios.
However, it would be a better idea to include something on SSM support
in the document (the least is to mention that SSM is not part of the
scenario and atmost explain what differentiates this solution from the
SSM scenario) since this document talks about multicast filtering rules
in IPv6. If the general consensus is that, i can work on including that.
I would like to see the stack doing filtering if necessary, so that when
a host joins specific sources (include mode) other packets reaching the
host from other sources would not be delivered to the application. And
also when in exclude mode, excluding specific sources, the stack should
filter in my opinion.
Yes. This is indeed SSM specific (when you want to exclude sources etc).
One way to include SSM in the draft is to talk about all this in the
filtering scenario, still suggesting that this is optional and not
default behavior.
In my opinion, the filtering should at least be done so that an
application doesn't receive from unwanted sources on the socket where
the filter is applied. If the same application or another, creates
another socket and binds to the right port and wildcard address or the
group address, then I think the best might be to deliver the packets.
I think this is what we present in the draft. If another application
binds to the right port and wild card address, it would still be
delivered the packet.
In my opinion, i guess it could be a good idea to talk about SSM in the
draft, in the filtering behavior section. I can even work on a separate
section that talks about Multicast Filtering in SSM based scenarios.
Do let me know what you suggest.
Thanks,
Arun T
Stig
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
