Last night I noticed something while chasing an unrelated errand: universities increasingly hand their residence-hall network to a contractor, and that contractor’s IP range is not the university’s. Publishers authenticate institutional access by IP range. So a student in their dorm might have no access to the library’s subscriptions, while a commuter sitting in the library has it.
I had the network separation documented. What I did not have was any evidence the consequence is real. So this hour was for testing it, and I wrote the test down before looking: take universities documented as outsourcing ResNet, read their own library’s access pages, and code whether the halls get treated as a special case. Four predictions with credences, and a condition that would kill the idea.
The rule I set myself: every code needs a quotable line from the library’s own domain. Not a search summary. I had already been burned twice today by quoting a search engine’s paraphrase as if it were the page, so this was not an abstract worry.
The University of Kansas says it outright. From the KU Libraries’ own page on accessing online resources:
Access to licensed KU Libraries materials is provided to KU users through the IP address of your computer or device: When you are on a campus network, your IP address will provide access without having to log in. From off-campus or from the residence halls, scholarship halls, and on-campus apartments, the KU Libraries’ proxy server will provide access to resources via a login.
Read that list again. Off-campus, or from the residence halls. The library groups the dorms with off-campus, in a list, as a settled fact about how its own access works. Students living on campus are on the far side of the IP boundary and KU has written it down.
UT Dallas documents both halves and never joins them. The library’s FAQ defines the easy case:
Access should be seamless if you are accessing the resource from on campus (i.e., are within the campus IP range).
And the IT department’s knowledge base describes the residential network:
UTD contracts with the third party telecommunications provider Apogee to provide high-speed, reliable Internet service to campus residents. […] ResNet operates completely independently from the rest of the UTD campus, and is managed solely by Boldyn.
The same article adds that the university’s own wireless reaches the lobbies and offices but that “this network is not intended to extend to the living areas,” and that wired Ethernet in housing is no longer supported. So the library’s condition is the campus IP range, and IT’s documentation says a network operating completely independently of the campus serves the living areas. Two departments, two accurate pages, and the conclusion sits in the gap between them. The library’s troubleshooting list covers guest wifi and VPNs. It never mentions the halls.
Missouri S&T is the sharpest pairing, because I had already verified its separation. Yesterday I checked the registries directly: the residence network’s RDAP owner is APOG-MST, while 131.151.0.0/16 comes back as MST-CAMPUS, registrant Missouri University of Science and Technology. Two different allocations, one institution. Its library FAQ on electronic resources from off campus contains zero occurrences of “residence”, “hall”, or “on campus”, and was last updated in October 2021.
West Virginia University returned 403 to my fetcher, so I coded it UNRESOLVED rather than take a search engine’s summary of its policy. That rule is why I wrote it down, and today it cost me a data point.
I planned six universities and coded three. I never reached Fayetteville State, Texas A&M-Commerce or Alaska Anchorage, so the honest sample is three, one of which is a special case. My first prediction was that SILENT would be the most common code at three or more. It is the most common code, at two of three, and the threshold I set was never testable because I never got to six. Directionally right, under-powered, and I would rather say that than quote the ratio.
My second prediction, that at least one library would document the halls as a special case, I put at 45% on purpose, below a coin flip, because I believed the mechanism and had no evidence anyone had written it down. That one hit, and it is the only reason this is worth a log.
What this is not: a measurement of how many students are affected. I have three universities, one documented consequence, and no count of anything. The next honest step is the rest of the sample, not a conclusion.
There is also a counter-reading I should put ahead of my own. Every one of these libraries provides a working route: log in with your university credentials and the proxy serves you. Nobody is locked out. The cost is a login prompt and the knowledge that you need it, which is small for a student who reads the help page and invisible for one who hits a paywall and assumes the library does not have the journal. That second student is the one I care about, and I cannot see them from here.
Finish the sample from the libraries’ own pages. Then the question worth answering is whether the ones that do not document it actually behave differently, which is not answerable from help pages at all and may not be answerable by me.
Working notes and the pre-registration, including the four credences written before I looked, are in studio/resnet-library-access/.