[rpd] New Policy Proposal Received - "Multihoming not required for ASN (AFPUB-2019-ASN-DRAFT01)"

[Jaco Kroon]

Hi Fernando,

On 2019/03/28 15:44, Fernando Frediani wrote:
> Well yes that we all know works technically but that's not the only 
> point.
> Now a days most companies will not be able to assign sufficient 
> resources to a downstream customers so having their own PI space makes 
> it more flexible and they can have as many necessary for their full 
> operation.

I am aware of multiple customers that only have addresses from Afrinic 
but no ASN, and their upstreams simply point statics at them which they 
then advertise forward.  The customer can move ISP by having new ISP do 
same and old ISP dropping the static, or even multi-homing by having 
both point statics.

The motivation here for an ASN would be so that if a link fails, so does 
BGP and traffic re-routes.

> Other reasons are: in many places you may not have access to two 
> providers but you may still need to have the PI space for example to 
> serve the local area, to have stable addresses, so in the future when 
> more upstreams you are ready to add it to the operation.
Agreed, but then you have INTENTION to multi-home based on which you 
should be able to obtain ASN under current policies.
> Also you may still have one upstream but you may be able to connect to 
> a local Internet Exchange or even to a local partner via  Private 
> Interconection.

In which case you're multi-homing and qualify for an ASN under the 
current policy.

>
> The requirement for Multihoming for requesting an ASN have been 
> removed in other RIRs which in my view is a natural thing so I believe 
> this is something that is not needed now a day as many aspects are 
> different from when it was made in this way.

To be clear:  I'm not against providing everyone that wants an ASN with 
one.  We now have 32-bit ASNs so there are currently one for every IPv4 
address.  In IPv6 there is one for every single /40 prefix (considering 
that only a single /8 is currently available for use, or every /32 which 
is the smallest allocation an LIR gets given).  So I think there is 
adequate available.  When there were only 16-bit ASNs that was different.

Could you please explain the use-case where you're not multihoming where 
having an ASN is beneficial?

Kind Regards,
Jaco


>
> Regards
> Fernando
>
> On 28/03/2019 10:07, Jaco Kroon wrote:
>> Hi,
>>
>> If you're not multihoming your upstream can simply point a static 
>> route towards you.  No need for an ASN.
>>
>> Even if you're multihoming that's still possible, but in that case 
>> you're more likely to want the control that having an ASN gives you.
>>
>> Kind Regards,
>> Jaco
>>
>> On 2019/03/25 17:17, Fernando Frediani wrote:
>>> Hello
>>>
>>> I want to give my support to the proposal for "Multhoming not 
>>> required for ASN" (AFPUB-2019-ASN-DRAFT01).
>>>
>>> As the author states in the summary and justification now a days 
>>> requiring Multihoming is something obsolete and nothing something 
>>> really necessary.
>>>
>>> With the exhaustion of IPv4 it becomes increasing important, 
>>> specially for newer companies to have their own IPv4 addresses as 
>>> they will hardly be able to have any reasonable size of allocations 
>>> from upstream providers as it used to be before. As also mentioned 
>>> one may wishes to have their own allocation and stable addresses, 
>>> will not depend from upstream providers and be able to announce 
>>> their PI space.
>>>
>>> Fernando
>>>
>>>
>>> _______________________________________________
>>> RPD mailing list
>>> RPD at afrinic.net
>>> https://lists.afrinic.net/mailman/listinfo/rpd
>
> _______________________________________________
> RPD mailing list
> RPD at afrinic.net
> https://lists.afrinic.net/mailman/listinfo/rpd