• 0 Posts
  • 12 Comments
Joined 1 year ago
cake
Cake day: June 10th, 2023

help-circle



  • Consider PCLinuxOS. ‘PLOS’ has the same look and feel of the ent Linuxes, but

    • as a child of mageia/mandriva from mandrake and conectiva, it’s derivation from RH is super long ago so it’s closer to rhel5 for well-built well-tested tools.

    • it has maaaaassive lib/app support range, like Axel Rose’s vocal range compared to EL’s Bruce Springsteen. No stream or other crap shenanigans aside from etc/alternatives.

    • No systemd. Weird how startups are fast and reliable

    It can yum cron like a badass.

    Caveats:

    • if you liked building vagrants on mageia, you need to help them on pclos. They have no clue there, and the skillet seems to be fading fast.
    • people who support sysv startup are getting more lazy and ditching it.
    • people who support last week’s version of anything are no more prevalent in pclos, so there’s no magical fix for “10 second tom” devs here either.




  • You’re really close, yeah .

    But because like every layer is checksummed both in delivery AND when it’s installed, so you can easily validate a delivered file, and it’s all signed with signatures you can easily check, you can at least be assured that

    • what you installed is what that package delivered
    • which is what the authors wanted
    • and the package probably hasn’t been tampered with
    • even weeks after install

    the chance of problems should be reduced.

    Bonus1: with a proper repo config, you can check for updates so fast. It’s like the chocolatey windows repo but more formalized and usually vendor-maintained.

    Bonus2: bad upgrade? Enterprise packages on Linux (long description; trust me) can be reverse-installed over what’s there so you can back-revise or downgrade with almost no pain. It’s a good oh-no fix. At every point you can still validate that what is there should be there, according to hard signatures at every stage.

    Bonus3: grabbing os version 6.1 and upgrading to 6.5 OR just installing 6.5 fresh gives the same final content - files and services - when you’re done. (almost entirely) No cruft, since package installs (because of the locking below) just install over themselves in a way Linux people just accept and windows people may freak over.

    Linux bonus: Linux locks file differently; again, long description, so trust me or look it up. You can upgrade many files and services without stopping them, and then bounce a service or a host, so your patch-and-bounce process is fast, it happens after the upgrades, and is like 2 min or with systemd 3min.

    Ultimately

    • use packages for wayyyy easier, consistent, reliable, tested, quasi-roll-back-able updates that you can validate all the way down.
    • and still that SNMP connection to check content remotely. It’s so great.



  • It’s a signed archive of deployable files along with meta-data. Usually a cpio archive (which is similar to a tarball) with that extra signature wrapper and meta-data (which, itself, should be a list of files and checksums).

    A proper package can validate a project’s installation, either from the local database or from remote resources, at any time, which gives positive assurance that what is installed is what should be installed.

    As well, proper package info is exported by SNMP to be consolidated centrally and validate what is vs what should be installed at the group level.

    TL;DR? Like a tarball with tracking info, signatures, checksums, and top-to-bottom validation. If it’s a good package, anyway.