|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] iSCSI simplificationThe word "profiles" is leading to unproductive analogies to profiles that have been used in other standards contexts. Many of the cited examples of profiles in other standards contexts are efforts to clean up interoperability messes after the fact, leading one of our ADs to inform us that making the mess in the first place will not be permitted: > there seems to be a misunderstanding somewhere - I was trying to say > nicely that a IPS protocol that has to have profiles would not pass > the IESG but I guess I was too nice - Not only will it not pass the IESG - it's rather unlikely to get to WG Last Call, as your WG co-chairs have a duty not to Last Call a document in that state. In particular, the Fibre Channel profiles are (unfortunately) examples of what not to do. We need to use words other than "profiles" because people don't agree on what that means, and some of the ideas it covers are actually useful. When Julian says: > have them REPLACE the plethora or features we have now and will > contain a fixed set of capabilities this is headed in a useful direction. Rather than using the term "profiles", I would talk about subsets of the protocol and conformance requirements. Truth be told, I did encourage Mark to post a pointer to the spreadsheet, not so much to encourage work on that spreadsheet, but rather, as Mark says: > [...] it will help show the sheer number of optional features we > are faced with, and may help us prioritize what must stay in the > protocol, and what we could live without in the interest of > simplicity. I count over 100 optional items in that listing, a number that has at least one too many zeros. Matt's comment on Fibre Channel use of profiles is an opportunity to make an important point: > Why have two profiles? Just like in fibre channel, have one profile that > describes the parameters to be used to be interoperable with other devices > (would be your profile 0). The other profile is simply the iSCSI spec, where > everything is open and negotiable. The IETF approach to this is to bit-bucket everything not involved in interoperability. The result is that there isn't a need for a "profile 0" because everything outside it is either excised from the spec or has words like "MUST" or "MUST NOT" (cf. RFC 2119) attached to it to indicate what is required for interoperability. Thanks, --David --------------------------------------------------- David L. Black, Senior Technologist EMC Corporation, 42 South St., Hopkinton, MA 01748 +1 (508) 435-1000 x75140 FAX: +1 (508) 497-8500 black_david@emc.com Mobile: +1 (978) 394-7754 ---------------------------------------------------
Home Last updated: Tue Sep 04 01:04:24 2001 6315 messages in chronological order |