Re: [Roll] Trickle Reset vs Restart

[Abdussalam Baryun <[e-mail]>]

That depends on the Protocol/Standard using Trickel Algorithm.

Restart can be used. So if such protocol using this algorithm needs
Restart, then it can be used for protocol performance and usecases.
Reset is always needed by Trickel ( it is in its six Rules, section 4) that
is in the RFC6206 specification, and also Reset can be needed/used by the
protocol using the Trickel (mentined in section 5).

Overall, any protocol/standard using this RFC6206 algorithm must specify
what does it reacts when inconsistency is detected. That reaction can be
Restart algorithm.

On Thu, Nov 21, 2019 at 7:02 AM Rahul Jadhav <[e-mail]> wrote:

> During the IETF106 roll-session, we discussed resetting vs restarting
> Trickle. Pascal raised this in another mail as well.
>
> After checking RFC 6206 (Section 4.2), I found that there is only one
> notion i.e., to reset Trickle timer, which means going back to Imin and
> starting a new interval.
>

not one notion I think,

>
> There is no notion of restarting the current timer for the instantaneous
> value of I.
>
> This is what my understanding was and reset is all that is needed, IMO. I
> would like to be corrected if there is any use-case, where restarting the
> timer would be needed.
>

I think Restart is needed by RPL at some topology changes with
inconsistency, or needed at protocol configuration changes at runtime. We
don't forget that C6206 is so far from perfect (these words mentioned in
section 6.7), with some efficient results. However, it may need simple
restart or reconfiguring its parameters at cases that can be noticed by
RFC6206 operational considerations section.


> We may not need any clarifications on this particular issue. Or do you
> think we should still keep this as a note for implementors in the
> observations draft?
>

IMO, clarification in drafts is always good specially in a
standard/updating_standard draft.

Regards,
AB