Re: CDI Distribution Draft

["Lisa Amini" <aminil@us.ibm.com>]

 > Sent by:    owner-webi@equinix.com
 > To:    Lisa Amini/Watson/IBM@IBMUS
 > cc:    cdn@ops.ietf.org, webi@lists.equinix.com
 > Subject:    Re: CDI Distribution Draft
 >
 >
 >
 > At 09:31 AM 11/26/2001 -0500, Lisa Amini wrote:
 > >  > >       Specifically, a master Resource Group could be defined
which
 > >  > >       would be used as the channel to communicate which Resource

 > Groups
 > >  > >       to be added/withdrawn from service.  The CIG would then
 > establish
 > >  > >       channels to receive cache coherency messages for these
 > Resource
 > >  > >       Groups.
 > >  >
 > >  > Yah, this is a nice architecture point for an end-to-end CDN or
 > inter-CDN
 > >  > system. Good to mention it in your draft.
 > >  >
 > >  > However, it is out side of RUP scope. RUP defines the operation
 > within one
 > >  > channel. How multiple channels are tied together or communicated to
a
 > > cache
 > >  > is a separate matter, and can potentially use any one of many
 > solutions.
 > >
 > ^^^^^^^^^^^^^^^^^^^^^
 > >I wasn't aware of any standardized solutions for this -- can you provide
 > >some pointers?
 >
 > e.g., manual config, SDP, WSDL/UDDL, PNP, ...

I agree on the manual config, but I don't think sending SDP files for this
is a good idea.  WSDL/UDDI is a good mechanism for describing these groups,
but what is currently available does not encompass what we are doing. I
personally think PNP would be overkill, but I'll check into it...I'll followup
with a more concrete proposal.

 >
 > It's fine to define a master control channel for tieing together all "CDN
 > channels" -- I'm actually quite interested in defining one. I just think it
 > needs to be tackled as a separate building block from RUP.
 >
 > >The metadata is as listed in Section 5.5, item 8.  I believe some of these
 > >fields map well to RUP (such as the Content Set (Resource Group),
 > >Authoritive Source (Content Location)), but others such as the Accounting
 > >Format Accounting Type, Request Routing Type, Time Frame (not the time for
 > >which the object is valid, but instead the time for which distribution
 > >services are being requested), ...).  So, yes, this metadata is important
 > >for interoperability, but some is also specific to CI.
 > >
 > >Given the need for interoperability and the fact that the current RUP
 > >document appears to be shifting toward covering both intra- and inter-
 > >domain, I consider it reasonable to specify this info in the RUP document.
 > >However, if it does not, as long as the RUP document does not preclude
 > >metadata being communicated as associated with content or resource
 > >groups, then we could publish a CI-specific document to cover these
 > fields.
 >
 > sounds like the right approach. Could the CDI Distribution reflect the
 > above comment / clarification?
 >

Yes, I'll clarify in the draft.

 > Thanks!
 > Dan
 >