Re: [Roll] Benjamin Kaduk's Discuss on draft-ietf-roll-turnon-rfc8138-14: (with DISCUSS and COMMENT)

["Pascal Thubert (pthubert)" <[e-mail]>]

Hello Michael 

I agree that for some link layer compression is not useful. I expect that this means higher power devices on less constrained links. For those there is configuration and management that can override the default . 

For unmanaged high volumes of low power devices and constrained radios, I’d rather have it on. And I agree that the node can only be a leaf in a mop 7 network.

Take care,

Pascal

> Le 12 sept. 2020 à 19:52, Michael Richardson <[e-mail]> a écrit :
> 
> 
> Pascal Thubert \(pthubert\) <pthubert=[e-mail]> wrote:
>> Now I’m unsure Michael and I agree anymore.
>> What will today’s developer code?
> 
>> We ask him to test if mop is < 7
>> What value will the developer place in his code if the test returns
>> false?
> 
> That's a good question.
> Pascal, it your intention that 8138 be turned on in this case?
> 
> I'm not opposed to that conclusion, but I thought that the goal was to move
> the signaling to mopex/capabilities/TBD?
> 
> Would there be cases in the future where ONLY 8138 compression would be coded?
> 
> We have set the situation where if a node does not understand the MOP, it can
> be a leaf, but 8138 incompatibility makes that not work.
> 
> 
>> Leaving it to the implementation will have some people choose true and
>> others false. This is not what we want.
> 
>> We want to control what the code does so we can expect it in the future
>> and build our backward compatibility based on that sure knowledge.
> 
>> Before the draft the default was no compression. Quite naturally since
>> initially it did not exist.
> 
>> Also we discussed on the ML that for RPLv2 all implementations MUST support the compression.
> 
> Just consider that for some link layers, 8138 compression might not make sense.
> 
>> In which case it is a better default for a coder today to decide to use
>> the compression for mop 7, isn’t it?
>> I hope I make the case right. Just think you’re coding it!
> 
> We're assuming someone coding today.
> Someone who is not writing MOP=7 code, has no mopex or capabilities code,
> because we haven't defined that.
> We agree that the "T" flag does not exist, and should be checked.
> Such a node would be a leaf.
> 
> I can live with defaulting 8138 to be on for MOP=7 for link layers for which
> 8138 is appropriate. (Someone coding today will know what link layers they
> have)
> 
> --
> Michael Richardson <[e-mail]>   . o O ( IPv6 IøT consulting )
>           Sandelman Software Works Inc, Ottawa and Worldwide