On Fri, Jul 29, 2016 at 5:49 PM, Jesse Gross <[email protected]> wrote:
>
>> On Jul 29, 2016, at 5:16 PM, Tom Herbert <[email protected]> wrote:
>>
>>> The only thing that I can say is that over the past several years since the 
>>> protocol was defined our experience with this tradeoff has been pretty 
>>> good. Both the number of uses of Geneve and implementations have increased 
>>> and as time has gone on, the uses have take more advantage of the 
>>> flexibility and have not run into any significant implementation issues 
>>> (either software or hardware).
>>
>> Jesse,
>>
>> In the list of TLV definitions you sent out I could only discern two
>> actual TLVs that have been defined. Cilium and VmWare stuff are
>> marketing slides, and an offhand reference to Geneve TLV in RCO draft
>> is by no means a specification. For comparison GUE has six defined
>> formally in I-Ds. So unless you're holding out a whole bunch of
>> development, I'm not seeing that the "inability to have a significant
>> number of extensions" is a material issue with GUE.
>
> Tom, I think that we’re going in circles here. All of the references I gave 
> with the exception of remote checksum have implementations behind them. Not 
> all of them are public. You can do as you wish with that information and the 
> other examples I gave.
>
> I think one practical difference between Geneve and GUE is that Geneve has 
> implementations and option definitions by a number of different parties. To 
> the best of my knowledge, all of the work on GUE has been done with your 
> involvement which makes it much easier to manage for your specific use cases. 
> I suspect that if GUE were to become a standard, you would find your 
> flexibility in this regard would be significantly reduced and might run into 
> some of the same issues as I’ve been describing. However, only time will tell.

I think it's equally possible that if Geneve becomes a standard you
will wind up with a mess of vendor specific options that overlap in
functionality but don't have any interoperability except between each
vendors devices-- great for vendors, miserable for users. You might
end up wishing that extensibility in Geneve had been limited in the
first place. Only time will tell...

Anyway, I suppose we can agree that extensibility is a strong
requirement, but we'll have to agree to disagree on what form
extensibility should take and how much an encapsulation protocol
should allow!

Tom

_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to