Re: [radext] Alissa Cooper's Discuss on draft-ietf-radext-ip-port-radius-ext-11: (with DISCUSS and COMMENT)

[Alan DeKok <[e-mail]>]

> On Oct 14, 2016, at 11:16 AM, <[e-mail]> <[e-mail]> wrote:
> Can you elaborate a little bit?
> What do you call 'appropriate parent TLV'?

  The idea is that if you have a TLV, it can:

a) define it's own sub-TLVs

b) use a pre-defined registry for sub-TLVs.

  That's it.

  If a TLV is defined to use option (a), there's no conflict with other TLVs.  It's sub-TLVs are completely independent of all other sub-TLVs.  And it *cannot* use TLVs from a registry.

  If a TLV is defined to use option (b), then it *cannot* define it's own sub-TLVs.  It MUST use the registry.  Which means re-using sub-TLVs from the registry.  Because that's what registries are for.

  sub-TLVs defined in a registry can't conflict with other sub-TLVs, because each TLV space is entirely independent... unless you define them to overlap.

> If a registry is defined to track allocated value for sub-TLV, how to avoid collision if "All other TLV parents are free to specify whatever sub-TLVs they want".

  That question is based on assumptions I don't understand.  As such, it can't really be answered other than with "it doesn't work like that".

  If we have a registry, fine.  Use it.  If we don't have a registry, then attributes not using the registry... don't use the registry.  And there's no conflict.  Because they're not using the registry.

  There is no more conflict here than with Service-Type and Framed-Protocol.  Both are "integer" attributes.  Both define a registry.  But adding an entry to Service-Type doesn't mean that that number is now also defined for Framed-Protocol, much less have any special meaning for any other "integer" attribute.

  Perhaps you could explain why you think there would be any conflict between sub-TLVs?  

  Alan DeKok.