Re: [CDNi] Fwd: New Version Notification for draft-warshavsky-private-features-metadata-00.txt

[Arnon Warshavsky <[e-mail]>]

Hi all
I was unable to attend IETF119 and I would like to relate to a concern re
this document that was brought up by Kevin Ma as for why do we need private
features rather than just create genericMetadata objects as we wish.

When a dCDN is digesting a configuration coming from a uCDN (same
configuration sent to several dCDNs),
every genericMetadata object can be handled in one of 3 ways:
1. digesting a known object registered either in SVTA schema or the dCDN
own private schema (and maybe rejecting it if not properly set according to
the schema)
2. rejecting an unknown object not appearing in any schema (e.g, a typo of
MI.Sorce instead of MI.Source would be rejected)
3. skipping a private object of a different dCDN which is identified as such

The structure of the private features of having a vendor scope alongside
the feature name,
is what enables 2 dCDN vendors define their own private features, and still
maintain bullets 1,2 intact.

e.g assuming we have in Qwilt a feature called*
'*MI.PrivateFeature.Qwilt.RunPythonScript'
that we will digest,
at the same time we will be able to skip
'MI.PrivateFeature.Lumen.RunLuaScript' appearing in the same configuration.

I hope this answers the question

thanks
/Arnon