Re: [Dime] OVLI: comments to 4.1

[Ben Campbell <[e-mail]>]

On Dec 9, 2013, at 9:56 AM, Steve Donovan <[e-mail]> wrote:

> Ulrich,
> 
> Looking at the impacts of removing the sequence number from the feature vector AVP.  This would require the following logic for each request received at the reporting node:
> 
> Parse the full set of OC-Feature-Vector AVPs 
> Calculate the set of overlapping capabilities
> Generate the reporting nodes OC-Feature-Vector AVPs
> If the reporting node is in overload then generate and append the OC-OLR AVPs.
> 
> Keeping the version number requires the reporting node to do the following:
> 
> Parse the feature vector sequence number
> If the sequence number is changed from the last version number received then 
>   Parse the remaining OC-Feature-Vector AVPs
>   Calculate the set of overlapping capabilities
>   Generate the reporting nodes capability AVPs
>   Save the version number and results of capabilities exchange
> If the reporting node is in overload then generate and append the OC-OLR AVPs.
> 
> Removing the sequence number forces the reporting node to do extra work for all requests, including requests received when overloaded.  Given that the sequence number will almost never change, this seems like a bad idea.  We should be reducing the work of overloaded nodes, not increasing it.

It's worth pointing out that, if someone believes the sequence number comparison is not worth the trouble, they can always ignore it and revert to Steve's first example.