Hi Keith! > -----Original Message----- > From: ietf <ietf-bounces@xxxxxxxx> On Behalf Of Keith Moore > Sent: Saturday, December 5, 2020 1:32 AM > To: ietf@xxxxxxxx > Subject: Re: TLS access Re: Call for Community Feedback: Retiring IETF FTP > Service > > Again, I find this analysis flawed to the point that it's useless or nearly so. > > The first rule in reporting stats, at least if you want them to be taken seriously, > is to describe the methods and results separately, and separately from any > conclusions. I have noted your concern and objection in the working summary. See https://github.com/ietf/ftp-retirement-consult-2020/commit/3f96335fc213424288eafd212ff925dd809f790b > The sampling period is far too short to be meaningful. If there were only 91 > unique users for a whole year, that's very different than only > 91 unique users over 12 days. But 91 users (if they're IETF > participants) would be a significant fraction of active IETF participants. > > Relationship between FTP and HTTP traffic volume seems irrelevant. Again, > traffic volume is not an indicator of importance. So why keep citing it? I keep citing the HTTP numbers because they provide context to the FTP numbers. I will say upfront that I don't know the entire population size of "active IETF participants". However, if we say that all 91, or even 140, IP addresses seen in FTP (let's include the search engines) are active IETF users. I see no reason why not to also count the >500,000 IP addresses that used HTTP as active IETF participants (see Section 2 of the "12 Days in the life ..." charts). Comparing 140 to >500,000 doesn't seem to align with calling this number a "significant fraction" (as its 0.02%). Regards, Roman > The conclusions made e.g. "99% of all FTP traffic.." are almost certainly not > warranted by the raw data, but the raw data aren't reported (not even > anonymized, not even in summary form). > > Keith >
