|
Hi Larry! Per “~91 different users that's a lot”, I’d like to put that into perspective – this number means that 140 unique IP address used the FTP service. In comparison, _only one_ of our HTTPS services exposing only a subset of this information
saw on average ~654,430 unique addresses. This makes FTP at best 0.02% of unique visitors (again, likely a lot less if full HTTPS logs were available). I don’t mean to trivialize the visits of these 140 IPs (minus the 35 that came from a search engine looking for robots.txt), but let’s be clear, there is a very, very, very small population being served with FTP as measured by request volume
(<0.2%) or unique visitors (<0.02%). See [1] for details. No access to information is being taken away. As noted in [2], alternative means for access the same information is available. Furthermore, analysis of the usage patterns [1] suggests that most
of the requests (96-98%) are trying to sync parts of the FTP repo. This style of access can be replicated with rsync. Hence, not only is access preserved, but also possible without custom tooling. Bottom line (for me) is that the data overwhelming shows that the community has moved to using other services. This is why the Tools Team asked the IESG to pose the question to the community.
I fully recognize that there may be custom scripts among those IP addresses hitting FTP. It isn’t clear how many, but it is no more than half of the 140 IP addresses [1]. Sunsetting FTP will break some code – those users will incur a
penalty here. To the community -- If you can’t or don’t have time to convert your FTP script; or need help configuring rsync, please reach out directly to me. I, volunteer developers from my employer (Carnegie Mellon University’s Software Engineering
Institute) or the Tools Team, within reason, will help you port your code or workflow, or answer questions about the services provided by the IETF infrastructure. Lastly, I’ve been sharing all sorts of updated stats in response to questions. I want to acknowledge and thank Glen Barney, Henrik Levkowetz, and Robert Sparks on the Tools Team who are pulling these numbers for us (found in [1] and [2])
and clarifying infrastructure details so we’re not speculating about hypotheticals. Regards, Roman [1]
https://docs.google.com/document/d/1JAXspeaMWFl8ML3hSezFSM0VsJsHI4uyDlQ2dHip8jo/edit# [2]
https://www.ietf.org/media/documents/Retiring_IETF_FTP_Service.pdf [3]
https://mailarchive.ietf.org/arch/msg/ietf/py_9b486x8x2io6d5dAb3FAgNng/ From: Larry Masinter <LMM@xxxxxxx> I read your data differently. If you have ~91 different users that's a lot. [Roman] That is for the community to discuss. Speaking personally, at first blush, perhaps 91 might seem like a lot.
We might have to agree to disagree on this. On the face it, perhaps 91 is a lot,
I agree with the sentiment that you've spent more attention bandwidth trying to retire the service than it should take to run it a few more years. Chrome had a 2-year plan to remove ftp from the browser by 2q 2021 Anyway, cool url's don't change. but c [2] |
