Here is the reply to the ops-dir review I just talked about.

Grüße, Carsten


Begin forwarded message:

> Resent-From: [email protected]
> From: Carsten Bormann <[email protected]>
> Subject: Re: Operations Directorate Review of draft-ietf-core-coap-13.txt
> Date: March 11, 2013 22:24:04 +0100
> Resent-To: [email protected], [email protected], [email protected], 
> [email protected]
> To: "Ersue, Mehmet (NSN - DE/Munich)" <[email protected]>
> Cc: "[email protected]" <[email protected]>, 
> "[email protected]" <[email protected]>, 
> ext Benoit Claise <[email protected]>, "[email protected]" 
> <[email protected]>
> 
> Hi Mehmet,
> 
> thank you for this input.
> 
> I read some of it as guidance for future development of the protocol, and in 
> particular for shaping the next steps the working group needs to take on; it 
> will be quite useful for this even if it is not influencing the base spec at 
> this time.
> 
> I believe we have to address scalability of the environments using the 
> protocol in separate work; how we will scale those up is somewhat orthogonal 
> to the base specification (which, however, is indeed intended to be 
> applicable to trillions of nodes).  Similarly, there are multiple models for 
> the interaction with other protocols such as DNS, and scalability will have 
> to be considered in refining these models.
> 
> NSTART is defined as 1 with a MUST; section 4.7 is specific about the cases 
> where different values will be allowed.  Again, defining congestion control 
> schemes more flexible than the base spec is future work.
> 
> Some specific comments:
> 
>> - The table in Section 4.8 needs a table number and title.
> 
> Fixed in SVN (will be in -14).
> 
>> - Section 4.8.1. "Changing The Parameters" says: "It is recommended that an 
>> application environment use consistent values for these parameters." 
>> However, it is unclear what the definition of "consistent" is in this 
>> context. The values in the table could be seen as an example of consistent 
>> set of variables. I would like to suggest to provide "value intervals" for 
>> the parameters, within which the entire parameter set appears to be 
>> consistent. Also it is unclear what is meant by "congestion-safe in the 
>> Internet" or "congestion control safety". Depending on this understanding 
>> the user of the protocol will be able to choose the correct values.
> 
> One problem is that it will require some ongoing research to answer these 
> questions.
> So we can't lock them down yet, and instead can just refer to general 
> principles.
> This includes limits on the values.
> 
>> - Concerning default values RFC 5706 states: "It is very useful to 
>> understand whether defaults are based on best current practice and are 
>> expected to change as technologies advance".
>> In the CoAP draft this has been done only for a few cases and could be 
>> further extended.
> 
> The current draft is quite clear that advances in congestion control 
> mechanisms are going to make different values possible.  Those 
> implementations that do not follow those advances are served reasonably well 
> by the standard values.
> 
>> - Section 4.8.2.  "Time Values derived from Transmission Parameters": I 
>> would like to suggest to summarize the discussed time values in a table with 
>> following column titles: Name, recommended default value, allowed value 
>> interval, error code if relevant.
> 
> We don't have the columns 3 and 4 yet.
> I have added a table with columns 1 and 2.
> 
>> ** Concerning the migration path, and future versions of CoAP in the same 
>> network:
>> 
>> - It would be useful to state clearly, in which cases it is dangerous to 
>> change any of the recommended default values or parameters in future 
>> versions of CoAP and would potentially impact the co-existence of two CoAP 
>> versions. Thus such a statement would support an incremental deployment to 
>> be successful.
> 
> Again, I believe this is future work, which also applies for the 
> configuration and management data model.
> 
> Protocol debugging is aided by the self-describing nature of the protocol 
> messages.
> This has worked pretty well in the (informal and formal) interops so far.
> 
> Future work will have to address management interfaces for CoAP nodes, 
> including management of security related events.
> I think some of this will have to integrate with resource discovery, an 
> active subject in the working group.
> 
> Grüße, Carsten
> 
> 
> 

_______________________________________________
Gen-art mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/gen-art

Reply via email to