Harald, Ok, there's some text in Section 7.5 that's already headed in that direction, so we'll see about writing a "MUST implement" requirement for the Counter64 items based on interface speed. Thanks, --David ---------------------------------------------------- David L. Black, Senior Technologist EMC Corporation, 176 South St., Hopkinton, MA 01748 +1 (508) 293-7953 FAX: +1 (508) 293-7786 black_david@emc.com Mobile: +1 (978) 394-7754 ---------------------------------------------------- > -----Original Message----- > From: gen-art-bounces@ietf.org > [mailto:gen-art-bounces@ietf.org] On Behalf Of Harald Tveit Alvestrand > Sent: Tuesday, January 17, 2006 11:17 AM > To: Black, David; gen-art@ietf.org > Cc: mbakke@cisco.com; marjorie_krueger@hp.com; > mankin@psg.com; yaronled@bezeqint.net; ips@ietf.org; > michele@sanrad.com; kzm@cisco.com > Subject: [Gen-art] RE: Gen-ART review of draft-ietf-ips-scsi-mib-08 > > Thanks for the quick feedback, David! > > I'm happy to leave this in your hands - one comment only: > > --On tirsdag, januar 17, 2006 11:02:36 -0500 > Black_David@emc.com wrote: > > >> The term "running at high speed" is a gating criterion for whether or > >> not the HS counters are mandatory, but I can't see that it's defined in > >> a testable way. Might have missed it - it would logically seem to > >> belong in section 7.5. > > > > Unfortunately, it's fuzzy and not testable in all cases. Here's what > > RFC 4181 (Section 4.6.1.2) has to say about this issue: > > > > Henceforth "standard" MIB modules MAY > > use the Counter64 type when it makes sense to do so, and MUST use > > Counter64 if the information being modelled would wrap in less than > > one hour if the Counter32 type was used instead. > > > > It clearly "makes sense" to use the Counter64 type, as there are SCSI > > implementations that clearly need it based on the "would wrap in less > > than one hour" criterion. Would adapting the quoted RFC 4181 text > > (with a reference to RFC 4181) be sufficient to satisfy your concern? > > What I'd like to see is something that makes it a complete no-brainer > whether or not the HC counters are needed, for instance: > > If the interconnect speed is higher than 4 Gbits/second, the HC counters > MUST be implemented, since that makes it possible to spin the counters > in one hour (see [RFC4181]). > > I wouldn't like someone to say "but... my implementation has a 10G > interface, but it's so badly implemented that I can't possibly get more > than 1 million operations per second through it, so I don't need to > implement the HC counters, do I?" > > (4G is picked out of thin air, but illustrates the problem... if The Number > is 3G, then 4G FC needs to implement it; if The Number is 9G, then only > people with 10GE and Infiniband interfaces need bother...) > > But you know this stuff, I don't.... > > Harald > > > > > _______________________________________________ > Gen-art mailing list > Gen-art@ietf.org > https://www1.ietf.org/mailman/listinfo/gen-art > _______________________________________________ Ips mailing list Ips@ietf.org https://www1.ietf.org/mailman/listinfo/ips