Re: [IPFIX] [GROW] [OPSAWG] WG adoption poll for draft-li-opsawg-ipfix-bgp-community-02

[PJ Aitken <[e-mail]>]

Stewart: yes, there could be multiple templates with multiple data records.

However there's nothing to provide the necessary "All of <these>" 
semantics to allow the collector to join up the lists in a meaningful 
way - ie, specifying which lists to use / not to use, and in which order 
if that's important.

Also, the data held by the collector may be indeterminate as the data 
records change - eg if list#1 and list#4 were updated, but updates 
hadn't yet been received for list#2 and list#3. So there'd have to be a 
"hold" time when nothing is changing and/or a signal that the current 
state is good - which then leads to questions of how to handle data loss 
(eg, if the update for list#2 wasn't received in time, or arrived late.

If single packet data export is no longer sufficient - as seems likely - 
then, since we already have several ideas, it'd be worth some IPFIX 
experts pondering what the correct solution should be.

P.


> On 16/02/2017 11:15, PJ Aitken wrote:
>>
>> IPFIX Collectors might reasonably assume that information in 
>> subsequent messages supersedes information in earlier messages, so 
>> splitting a list across multiple messages would not have the desired 
>> effect.
>>
>>
>
> Hi Paul
>
> Could you not have multiple discrete templates?
>
> Stewart