The taxonomy defines 'surrogate origin server', which is what AFAIK most
CDNs operate as. It also defines several mechanisms for inter-replica
communication (i.e, between the master origin server and surrogates). Most
CDNs will be using some form of demand-driven or batch replication - IMHO
these definitions are general enough to catch current implementations.
On Tue, Feb 15, 2000 at 03:23:24PM +0100, LOVRIC Ivan CNET/DMI/CAE wrote:
> It is the goal of the taxonomy to describe the current practice related to
> cache and replication. IMO, CDNs should already be considered as a current
> practice, though they use proprietary protocols; so they are in the scope of
> the taxonomy.
> Nevertheless, I agree with you : this item did not exist in the first half
> of 1999, and it should be mentionned only in a next version of the taxonomy,
> if the WG wants to work on that topic.
>
>
> -----Message d'origine-----
> De: Gary Tomlinson [mailto:GARYT@novell.com]
> Date: mercredi 9 février 2000 21:51
> À: ivan.lovric@cnet.francetelecom.fr; ian@mirror-image.com
> Cc: wrec@cs.utk.edu
> Objet: Re: Taxonomy about CDNs
>
>
> I agree with Ian's assessment that CDN's are little out of scope for the
> Taxonomy. My reasoning here is, the ones's I'm familair with use a
> combination of open protocols and some proprietary protocols and algorithms
> to make them work. The CDN's I've worked with gaurd thier "secret sauce"
> proprietary value adds as they see these are core differentiators today.
> So, I would find it hard to document them in taxonomy given the proprietary
> constraints.
>
> The whole concept of the taxonomy is a little difficult to capture
> accurately at any point in time. We've done a good job of capturing the
> caching and replication space as it existed in the first half of 1999. In
> order to get the document published, we need to have a sunset on it. This
> of course leads to the need for WREC to define its future with respect to
> taxonomies. We may want to consider producing a new version over the next
> year that emcompasses the last 12-18 months in this space, including CDN's
> (with help from some of the CDN's).
>
> Gary
>
> >>> Ian Cooper <ian@mirror-image.com> 02/09/00 10:12AM >>>
> On Tue, 8 Feb 2000 14:23:02 +0100 , LOVRIC Ivan CNET/DMI/CAE wrote:
>
> >For the next IETF, I suggest to add CDN considerations to the taxonomy
> >draft.
> >
> >Vendors like Akamai, Adero, Sandpiper use caches and/or mirrors to deliver
> >content.
> >At present, it is still difficult to see if CDN vendors use proprietary or
> >standard protocols. Also, it is not easy to define exactly what are the
> >boundaries of content delivery.
> >That is why, IMO, CDNs are in the scope of the WREC WG, and therefore
> >appropriate taxonomy should be defined.
>
> Do you mean the taxonomy? (Or am I just totally brain dead today?)
>
> Section 5 of the taxonomy describes inter-replica communication (I
> need to go back and see what wording it uses to check that we're
> covered for things like "Master Origin Server"), and 3.1.1 describes
> client-replica communication.
>
>
> While I agree that the discussion of CDNs, and the protocols that are
> used to distribute content among replicas, is useful, I'm not sure
> that this belongs in the taxonomy itself.
-- Mark Nottingham, Senior Developer Akamai Technologies (San Mateo, CA)
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:27 MST