We're a bit reluctant to respond to this, in part because the matter is a lot more nuanced than it might appear. We had an internal debate about whether any press reporting of it would adequately reflect that, and whether we would be better served by simply publishing something on our web page. Ultimately, I won the argument on the basis that your audience aligns with the people who're asking about this, and responding might reduce the number of times we get asked the same questions. I'm hoping I don't regret that.
With that disclaimer, here are the answers to your questions:
When was mirror.ac.za officially decommissioned?
The official end date was 10 September 2025. However, in reality, it was shut down a bit after that.
Well before the planned decommissioning date, we tried to make contact with
every project for which we were listed as an official mirror and requested that they remove us from their mirror list. We only decommissioned after we'd received confirmation from a project. Unfortunately, with some hindsight, some projects confirmed they had removed us (or would before the end date), but haven't actually cleaned up properly. For example, we contacted the FreeBSD Project during August and received confirmation from them on 19 August. However, as of a few minutes ago,
ftp.za.freebsd.org still points to our non-existent mirror.
It's for that reason that we've created placeholder websites for every single decommissioned mirror. We deliberately return an HTTP 410 Gone response so that search engines and other crawlers will de-list the page, and end users will receive a valid error response. In most cases, the error message will refer you to the Project's official mirror list. And where we know of viable local alternatives, we've listed those too (see, for example,
https://freebsd.mirror.ac.za/). We'll maintain those placeholders for at least six months, although the non *.mirror.ac.za ones will get certificate warnings once their existing SSL certificates expire.
Why was it decided to decommission TENET's mirror service?
The short answer is that we no longer have a mandate to continue.
To properly explain that, we need a bit of history and an explanation of how TENET works.
TENET is a non-profit company that was set up specifically for the benefit of South African (public) universities and associated research institutions. That's the community we serve, who funds us, and who we're ultimately accountable to. We serve at their behest, and that's where we get our mandate.
The mirror service was created in about 2012, at a time of relative bandwidth scarcity and complex interconnection requirements. At the time, the motivation for creating it revolved around bandwidth and peering arrangements. But importantly, being a community-led organisation, the service was governed by an advisory group drawn from our client institutions, who made decisions about what projects should be mirrored and why. One of the decisions at the time was that the service should be generally accessible and, wherever possible, we should be an official mirror for any project we hosted. That led to users of the service well beyond the community that TENET itself actually serves.
Fast forward, and South Africa has a lot more bandwidth. Additionally, there are more submarine cables, resulting in a more reliable Internet in general. Over the last few years, we've seen substantial changes in the usage patterns of the mirror service. For many of the projects we mirror, we had reached the point where the bandwidth expended in maintaining a mirror exceeded that of the clients. In other words, rather than saving bandwidth (its original justification), the service now had a net cost. The five-minute average of client traffic just before we decommissioned was less than 100Mbps (whereas the server has multiple 10Gbps links and used to average gigabits per second).
Where mirrors were seeing significant client traffic, the majority of that was not from our own community or even South Africa. The only mirrors where South African clients dominated were those where we were an official mirror behind a load balancer that did proper geolocation. For all the others, with one exception, African countries never appear among the top users. For example, we saw more USA and Dutch users of the Alpine, docker, and Ubuntu cloud & release mirrors than South African; Macports was more popular in Brazil; CPAN was used in Singapore and France; Raspbian was used in Cameroon and the Netherlands more than South Africa; FreeBSD is popular in Brazil (with ZA in 16th place); etc.
However, as telling as those numbers are, TENET has lots of bandwidth, and bandwidth is
not the reason this went away (if it were, we would have shut it down several years ago). Maintaining a large mirror service takes staff time, and it requires hardware, power, and other resources. So, at the point at which that hardware (and simultaneously, the operating system) reached end-of-life and we needed to spend money on replacing it, we began to consider whether this service was still adding value. And importantly, whether the time and resources would be better spent on other projects.
Ultimately, we do what the institutions we serve want (and are willing to pay for). If they wanted us to maintain a mirror service, we would. We didn't arbitrarily decide to shut the service down — we asked the community that pays our bills for input. However, that call was met with deafening silence. Indeed, all the queries we've received about shutting it down have come from outside the community that funds us. And perhaps most tellingly, the last active institutional member of the advisory group responsible for the service left in 2017, with several subsequent attempts to resurrect it having failed. What all of this means is we don't have a mandate from the institutions we serve to expend further resources on this.
And yes, we understand arguments about digital sovereignty, African hosting of international projects, etc. But as a small non-profit, we also have to be pragmatic. So what we're doing now is redirecting the resources we might have expended on this into digital sovereignty exercises that directly impact higher education, particularly around the FAIR principles and safeguarding South African research output. That's far more directly aligned with the mission of an NREN (national research and education network) than running a mirror service.
I got some great stories with that guy 