Re: [Tsv-art] Tsvart last call review of draft-ietf-ice-rfc5245bis-16

[Ben Campbell <[e-mail]>]

Commenting on the first issue:

> On Jan 30, 2018, at 8:34 AM, Christer Holmberg <[e-mail]> wrote:
> 
>> Significant Issues:
>> 
>> A. Section 5.2:
>> 
>>   Lite implementations only utilize host candidates.  A lite
>>   implementation MUST, for each component of each data stream, allocate
>>   zero or one IPv4 candidates.  It MAY allocate zero or more IPv6
>>   candidates, but no more than one per each IPv6 address utilized by
>>   the host.  Since there can be no more than one IPv4 candidate per
>>   component of each data stream, if an ICE agent has multiple IPv4
>>   addresses, it MUST choose one for allocating the candidate.  If a
>>   host is dual-stack, it is RECOMMENDED that it allocate one IPv4
>>   candidate and one global IPv6 address.  With the lite implementation,
>>   ICE cannot be used to dynamically choose amongst candidates.
>>   Therefore, including more than one candidate from a particular scope
>>   is NOT RECOMMENDED, since only a connectivity check can truly
>>   determine whether to use one address or the other.
>> 
>> I find it quite strange that the above text says there can only be single
>> IPv4
>> based candidate, while for IPv6 a LITE implementation may have one
>> candidate
>> per IPv6 address. Isn't the LITE implication of having multiple
>> candidates for
>> the same address family similar? Yes, IPv6 kind of forces the need for
>> dealing
>> with multiple IPv6 addresses on any host. However, I can see that certain
>> servers will actually be multi-homed in IPv4 and thus can in a sensible
>> way
>> actually have multiple IPv4 candidates, and let the clients select which
>> interface has the best reachability.
>> 
>> Can you please be explicit on what in ICE prevents things to work for
>> IPv4 but
>> the same case works for IPv6?
> 
> This is text from RFC 5245. I agree it is confusing, and unfortunately I
> don¡¯t have a good answer.
> 
> I guess my approach would be to suggest that we simply remove the
> restriction. In addition, there is generic text about dual-stack etc
> elsewhere,
> and I don¡¯t see anything ICE lite specific.
> 
> OLD:
> 
> "Lite implementations only utilize host candidates.  A lite
>   implementation MUST, for each component of each data stream, allocate
>   zero or one IPv4 candidates.  It MAY allocate zero or more IPv6
> candidates, but no more than one per each IPv6 address utilized by
>   the host.  Since there can be no more than one IPv4 candidate per
>   component of each data stream, if an ICE agent has multiple IPv4
>   addresses, it MUST choose one for allocating the candidate.  If a
>   host is dual-stack, it is RECOMMENDED that it allocate one IPv4
>   candidate and one global IPv6 address.  With the lite implementation,
>   ICE cannot be used to dynamically choose amongst candidates.
>   Therefore, including more than one candidate from a particular scope
>   is NOT RECOMMENDED, since only a connectivity check can truly
>   determine whether to use one address or the other."
> 
> 
> NEW:
> 
> "Lite implementations only utilize host candidates.
> With the lite implementation, ICE cannot be used to dynamically choose
> amongst candidates. Therefore, including more than one candidate from a
> particular IP address family is NOT RECOMMENDED, since only a connectivity
> check can Truly determine whether to use one address or the other.”
> 

We should avoid making non-critical changes to text that was unchanged from 5245 at this point in the process. If people think this is important to change, it needs working group discussion. This is a hazard of bis drafts—it’s really hard to tell when you are done. :-)

I am guessing that the original motivation was that, since ICE-lite cannot select among candidates, you want the minimum number of candidates necessary per address family. Multiple IPv6 candidates are allowed because of the nature of IPv6. But I think if you are truly multi-homed, you probably shouldn’t be using ICE-lite.

Ben.