Hi, On 2023-01-29 20:11, Gert Doering wrote:
Hi,
On Fri, Jan 27, 2023 at 02:15:44AM +0100, Robert Scheck wrote:
On Mon, 09 Jan 2023, Gert Doering wrote:
So don't do OS updates. But *do* update the probe software, automatically, and in normal intervals.
writing while wearing my hat as a RPM packager: I'm not really sure how this shall work practically: If the probe software is run on regular (non- hardware probe) systems, there are runtime dependencies to other software, e.g. due to dynamic linking. Latter is encouraged by Linux distributions, [..]
This is all good and true - so, do not distribute as RPM.
We produce the ("self signed") RPMs because we need them internally for the anchors anyway, so figured we may as well publish them, plus some people actually need/prefer binary RPMs. Otherwise we're heading to a direction where we act as the upstream for the software, and professional package maintainers do the proper signing, distribution and updates (like Robert S. could become one for RPMs).
Distribute as Docker image, or Flatpack, or whatever all these new "encapsulate a program and all its dependencies" approaches are called, and I'm sure some of them have mechanisms to enforce timely updates.
Docker files exist, though we are not maintaining official images. That may change in the future. As said before, "have mechanisms to enforce timely updates" indeed has proper solutions for basically all platforms.
First and foremost, a RIPE Atlas probe is a centrally managed piece of software, and this is what it needs to be: centrally managed, centrally upgraded, always at the upgrade level the central controller wants it to be.
As I tried to explain before, this is true for the HW probes, not so much for SW. Cheers, Robert
Gert Doering -- NetMaster