Re: [Dime] Conclusion for the Report Type

It's probably also worth adding some guidance on when it makes sense to define a new report type. For example, if I want to add a new AVP that can be safely ignored by a recipient that doesn't understand it, I probably don't need a new report type.

On Dec 9, 2013, at 6:00 AM, Jouni Korhonen <jouni.nospam@gmail.com> wrote:

> Folks,
> 
> We need a conclusion here so that I can actually write something
> into the -01. How about the following (I try to reflect as many
> points given here as possible):
> 
> 1) The basic principle for the Report Type use is that only one
>   OLR per report type is allowed unless the report type and the
>   OLR reflecting the new report type define exact semantics how
>   to differentiate between multiple OLRs with the same report
>   type. In 3GPP context, for example, a report type with an AVP
>   that identifies an APN could be such a differentiator.. and that
>   would need a new report type where an implementation exactly
>   knows to look for this additional AVP without guesswork or 
>   fuzzy heuristics.
> 
> 2) A new report type or a set of new report types require a new
>   feature to be allocated/defined so that both endpoints know how
>   to handle the new report type that was defined after the
>   publication of the baseline specification. The handling of the
>   new report types must be defined (along with the new AVPs it
>   might need to be included into the OC-OLR AVP).
> 
> 3) With 2) in place I do not care whether the OC-Report-Type is
>   enumerated or unsigned (flag vector?). I still favour Enumerated
>   myself as it forces the protocol designer to come up with a 
>   cleaner design ;)
> 
> 4) For the baseline we only define host and realm report types.
>   We do not allow multiple OLRs with these report types i.e.
>   single instances of OLRs with host and/or realm are allowed.
> 
> - Jouni
> _______________________________________________
> DiME mailing list
> DiME@ietf.org
> https://www.ietf.org/mailman/listinfo/dime