In determining customer requirements for "multihomed" or generic high-availability services, I often find there is a major terminology problem. Customer expectations may not be met because they are using "cache" or "mirror" differently than an ISP person. After the April IETF meeting, Jacob Palme posted to the general IETF mailing list what I tink is an excellent start on the problem. There was almost no response, other than a comment or two that different application communities use the terms differently. Perhaps these issues would be more of interest in NANOG, because I think they are very operational at the requirements definition end. Unless a more useful place to collect comments emerges, I'd like to record the consensus in my I-D on multihoming requirements, currently http://www.ietf.org/internet-drafts/draft-berkowitz-multirqmt-01.txt My thoughts would to be to get one more round of comments, see if the IETF Operations Area wants to consider it, and if not send it to the IESG as an individual contribution in a month or two. The subject is one of those things that doesn't quite fit the RFC model. ----------- Date: Thu, 2 Apr 1998 15:38:13 +0100 To: IETF general mailing list <ietf@ns.ietf.org> From: Jacob Palme <jpalme@dsv.su.se> Subject: Terminology for replication What is the correct term for various kinds of replication. Below is a proposal as a basis for a discussion. Function Proposed term -------------------------------------------------------+------------------- A general term for all kinds of copying of identical Replication to several stores. Replication, where the stores which hold the ori- Mirroring gal knows about where copies are held and periodic (I have also heard automatic updates ensure conformance between mirror the term push- copies and the original. caching) Replication, where a store saves a copy of a docu- Caching ment which it happens to get, without 100 % safe methods against outdated objects in the cache. Caching with some added facilities, such as Controlled caching? "Expiry" fields, to reduce the risk of outdated objects in a cache. Replication, where all copies are equally valid, ????? none of them is the "original", as in Usenet News. ------------------------------------------------------------------------ Jacob Palme <jpalme@dsv.su.se> (Stockholm University and KTH) for more info see URL: http://www.dsv.su.se/~jpalme