Hi Brian,

On 11/29/2016 11:43 AM, Brian E Carpenter wrote:
> Joe,
>
> On 29/11/2016 17:38, Joe Touch wrote:
>> Hi, Brian,
>>
>>
>> On 11/28/2016 7:59 PM, Brian E Carpenter wrote:
>>> Hi,
>>>
>>> My first question is not whether it's a good idea to build an IP VPN over
>>> IP tunnels, because I'm sure it is. It is more whether we actually need
>>> a BCP describing how to do it, rather than just, say, open-source code
>>> for a VRF instance that does this.
>> +1
>>> I think that question is definitely worth exploring, and is probably a big
>>> enough question to deserve a BOF (not necessarily a WG-forming BOF). But
>>> that needs to be based on a more problem-oriented and analytic draft, I 
>>> think.
>>> It definitely needs expertise from the Transport Area as well as the 
>>> Internet
>>> Area, to get the congestion management right.
>> -1
>>
>> We already have RFC6040.
> Doesn't that only apply with ECN-capable end points?

Yes. Where those aren't available, IP over IP tunnels (or IP over X over
IP) provide the same "congestion control" that all links do (i.e., none).

>
>> This isn't a transport problem (if it is, it
>> has been done incorrectly - see below).
> No, but it might induce interesting transport problems if the tunnel behaves
> other than as a piece of wire. RFC6077 seems to identify a number of open
> problems in this area, but you certainly know more about this than I do.

The place to go here if you want bandwidth guarantees or rate limits
within a tunnel is PWE, but again that's not transport. I'm not saying
there should be no congestion control, only that it's a different animal
than at the transport layer.

>
>>> For the moment, I am quite unable to judge whether the proposal in this 
>>> draft
>>> to use GRE-in-UDP or GUE is the best answer.
>> There can be no single answer to that question. Like regular links,
>> tunnels (virtual links) vary with their environment, and should.
>>>  I also don't really understand
>>> the security model. There is some discussion of IPsec tunnels and RFC3884.
>>> If we use IPsec tunnels, why would we need DTLS? For that matter, if we use
>>> TLS tunnels, why would we need DTLS?
>> TLS is a very bad idea. We should never try to tunnel IP over TCP.
> I agree it's a terrible idea, but pragmatically situations can arrive where 
> it's
> the only real option.

Sure, but it's bad enough to try to avoid if there are other alternatives.

>> DTLS might be available where IPsec isn't.
> If that is the case, it needs to be explained in more detail in the draft.

Agreed.

> Anyway it needs to be a clear choice: IPsec or DTLS, but not both.
> (This is a point that has also come up in the Anima WG, as it happens.)

Agreed as well.

Joe

>
>    Brian
>
>>> I'm also quite unable to know how to position this proposal compared to
>>> https://tools.ietf.org/html/draft-templin-aerolink which has been
>>> in development for several years. They seem to tackle some of the same
>>> problems.
>> +1
>>> Regards
>>>    Brian Carpenter
>>>
>>>
>>> _______________________________________________
>>> Int-area mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/int-area
>>

_______________________________________________
Int-area mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/int-area

Reply via email to