[ofa-general] Re: [PATCH] opensm: update Release Notes for OpenSM version 3.2

[Hal Rosenstock]

Sasha,

On Tue, Oct 14, 2008 at 5:21 AM, Sasha Khapyorsky <sashak at voltaire.com> wrote:
> Hi Hal,
>
> On 07:38 Sun 12 Oct     , Hal Rosenstock wrote:
>>
>> > +  When this mode is used (enabled with --consolidate_ipv6_snm_req option)
>> > +  OpenSM will map all IPv6 Solicited Node Multicast address join requests
>> > +  into a single Multicast group with address ff10:601b::1:ff00:0. In this
>> > +  way limited MLID space is saved. The feature is very useful with large
>> > +  (~> 1024 nodes) clusters.
>>
>> Isn't there truth in advertising ? This should also say:
>>
>> This IBA noncompliant feature may be problematic in heterogeneous
>> subnets (due to rate and MTU differences).
>
> The potential problem in heterogeneous subnets is not much different
> from IPv4 - 1x ports will not be able to join. Is it the case you meant?
>
> If yes, IPv6 SNM becomes much less critical issue comparing to not
> working IPv4.

Isn't it at best equal rather than much less critical ? The reason I
say at best equal is that it's the first joiner's parameters are used
for all SNM groups which could "lock out" the majority of users.
That's not the case with IPv4 where the group parameters are
configured and only those which don't match can't participate. There's
also a difference in that one is IBA compliant and the other is not.

So this doesn't warrant mentioning as far as you're concerned ?

-- Hal

> Sasha