Re: [v6ops] Discussion of draft-ietf-v6ops-ula-usage-recommendations

[joel jaeggli <[e-mail]>]

On 7/21/15 6:32 PM, Alexandru Petrescu wrote:
> 
> 
> Le 21/07/2015 18:28, Mark Andrews a écrit :
>>
>> In message <[e-mail]>, Alexandru Petrescu writes:
>>>
>>>
>>> Le 21/07/2015 16:53, Brian E Carpenter a crit :
>>>> On 22/07/2015 02:18, Alexandru Petrescu wrote:
>>>>> 1. Brian suggested to recommend that globals should be there on
>>>>> the machines having ULAs as well, if I understand correctly.
>>>>>
>>>>> But I think so only on some Hosts, mainly the Hosts of end users.
>>>>
>>>> All hosts that need external communication.
>>>
>>> I agree, all hosts that need external communication.
>>>
>>>
>>>>> 2. the ULA RFC suggests a ULA prefix can be generated out of a MAC
>>>>> address.  That sixxs implementation does it.  Except it takes it
>>>>> too serious: it does not accept a MAC address which is not a real
>>>>> MAC address - in that oui.txt.  And random MAC addresses (for
>>>>> privacy) certainly are not in that oui.txt.
>>>>>
>>>>> I think this is an undesirable situation to be in: unable to
>>>>> generate ULAs because the only tool out there (sixxs) can't refuses
>>>>> a copy paste a MAC address from the widely used windows 7 laptops.
>>>>
>>>> That isn't a standards issue, but I agree that operationally, there
>>>> needs to be a viable way for anyone to generate a random number. Wait
>>>> a minute, that doesn't seem hard.
>>>
>>> It's easily done centrally, but in a distributed manner it's harder -
>>> how am I sure the network I connect to has ULAs generated such that they
>>> dont clash with mine?
>>
>> *YOU* generate you ULA properly.
> 
> You for single or plural?
> 
> Until now I was saying to my peers: I take 192.168.1.1 please take
> something else and I'll route to you.

that's to of dramatically underselling the amount of coordination
required to use rc1918 space between administrative domains.

> Now I am saying: generate ULA properly, make it truly random.  But how?
> 
>>>>> I am not sure what the problem is, but it's very good to have a
>>>>> very easy way to generate ULAs.
>>>>>
>>>>> 3. in an enterprise deployment there was a problem of ULAs deployed
>>>>> in a intra-network and another ULA space in another intra-network,
>>>>> of the same enterprise.  So we wanted to make sure two things: the
>>>>> two ULA spaces are distinct, or otherwise make sure the gateway
>>>>> router does not route between the two intranets' ULAs (but yes,
>>>>> route between their respective GUAs).
>>>>
>>>> Why not? ULA to ULA routing on a private link might be desired (e.g.
>>>> after two networks merge without renumbering). From a routing PoV
>>>> there is nothing special about a ULA prefix; we just need to
>>>> configure carefully where it is routed and where it is not routed.
>>>
>>> Yes, private routing should be ok, but only if these ULAs are unique.
>>> If people on different networks use different generation methods then
>>> it's dubious to be sure of the uniqueness.  Maybe I choose fd00:1::/64
>>> being sure that no random generator will make it, and it happens my
>>> neighbors does the same.  That leads to conflict on fd00:1::/64 and we
>>> dont want routing enabled between the two.
>>
>> Generate.  Don't choose.  If you generate then you should be ok.
> 
> Ok.
> 
> But it's much easier to choose.
> 
> We want simple ULA addresses, simple to remember, simple to type,
> ideally based on a dictionary.

Wierdly when you go to an RIR you don't get to pick what your prefix is...

if you don't see your selection with randomness your probability of a
collision within your organzation is going to go way up.

> This is something everybody building even the simplest IPv6 network has
> to do: what simple ULA IPv6 addresses to put there to not break something.

I don't have any ula in in my production networks.

> Alex
> 
>>
>>>> Anyway - I'd like to see the draft progress. Has it already had a
>>>> WGLC?
>>>
>>> I agree, it already has advice in it worth progressing.
>>>
>>> Alex
>>>
>>>>
>>>> Brian
>>>>
>>>>> I am not sure how to translate that into advice, because I am not
>>>>> sure how it will unfold in the near future.
>>>>>
>>>>> Alex
>>>>>
>>>>> Le 21/07/2015 16:02, Fred Baker (fred) a crit :
>>>>>> https://tools.ietf.org/html/draft-ietf-v6ops-ula-usage-recommendations
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>>
>>> "Considerations For Using Unique Local Addresses", Bing Liu, Sheng
>>>>>> Jiang, 2015-05-03
>>>>>>
>>>>>> This draft came up from the floor this afternoon. I think we
>>>>>> need some concentrated constructive conversation regarding it -
>>>>>> we have had a lot of the other kind.
>>>>>>
>>>>>> What issues do we need to address to complete it. and what
>>>>>> specific recommendations would that include?
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________ v6ops mailing
>>>>>> list [e-mail] https://www.ietf.org/mailman/listinfo/v6ops
>>>>>>
>>>>>
>>>>> _______________________________________________ v6ops mailing list
>>>>> [e-mail] https://www.ietf.org/mailman/listinfo/v6ops
>>>>>
>>>>
>>>>
>>>
>>> _______________________________________________
>>> v6ops mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/v6ops
>>
> 
> _______________________________________________
> v6ops mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/v6ops
>