Russell -
Thanks for a very succinct and clear breakdown. I'm a BSD veteran who
has enjoyed OSX and Linux for their familiar basis, but am often
confounded by the never-ending neologisms that roll up and out of the
popular computing culture.
I've been involved in different generations of distributed computing
over the decades but definitely have not kept up with the terminology or
any specific technologies for years and years.
Recently glen referenced Web 2.0, 3.0, 4.0 in a way that illuminated my
lack of up-to-date-ness on the detailed popular terminologies... I
*think* of this progression as being about raising the level of
abstraction of what is being distributed in each of the generations?
With 4.0 being about distributed decision making, context awareness,
inference, action, etc.
I was inspired when I read Google's Matryoshka model outline:
https://arxiv.org/abs/2205.13147 for the possibilities for distributing
cognition assistance and human-AI collaboration at escalating levels of
resolution/fidelity.
I don't have a dog in the fight any longer (no professional projects in
this domain) but do have a "pet dog" which keeps wanting to mix it up
with the big dogs. I use LLMs across a range of personal (and "hobby")
purposes and modes and have my own paranoias about corporate and
government exploitation, surveillance and interference in such, so am
drawn to any ideas of it all being distributed effectively from edge to
cloud or cloud-of-clouds or better yet "mycelial" peer-to-peer mashups.
I've done various projects (now decades in the past) involving
multi-resolution modeling and federated (e.g. HLA) modeling and feel
that perhaps the ideas and tools of those early days are now becoming
somewhat mature? This business of ubiquitous "containers' as
"lightweight VMs" or contexts to run them in, and multiresolution LLM
architectures like Matryoshka and the very pop-term of Web 4.0 seems to
reflect such?
Guerin's "Acequia vs Cloud" schtick aligns well with this too... along
with P2P *lateral* or holonic/hetararchic vs hierarchic distributions.
On 4/18/26 10:02 pm, Russell Standish wrote:
This stuff is a whole big subject, so I'll just give a short replay
here, to help you decide wether its worth digging in more.
Systemd is essentially the second process on a Linux system, after the
kernel (ie Linux) has started. It is responsible for launching all
other processes after that, including all the "daemons" (ie services
such as web servers, databases and the GUI), and user programs like
web browsers, word processors and command line "shells", and
arbitrating between their competing resource requirements. Older
technology for doing that was called Init, and there is some elements
of controversy about whether systemd is a good thing or not.
Containers are essentially "cut-down" versions of Linux - you can
think of them as a light-weight virtual machine, they look and feel
like another Linux computer running on your computer, but actually
just run the processes on your host machine with restrictive access
premissions.
The main container technology is called Docker, although alternatives
exist (like most things). Typically, you would spin up a packaged
instance of a Linux OS as a container, then add whatever software you
want to run it. I don't think you can run GUI software (eg Wayland or
X), unlike the more heavyweight virtual machine technology such as
VirtualBox or VMWare. In particular, I've never seen Windows run in a
Docker container, but its easy to set up Windows in a Virtual
Machine. So oriented more for setting up servers, not
workstations. For GUI applications, they're typically implemented as a
Web application, so the Docker container provides a web server, and
the GUI is simply the normal web browser on your host computer.
AI agents is another whole box of wax. Here, the agents need to run
programs to effect their work - eg a software developer agent needs to
compile code and run regression tests, which can be easily set up on
containers to isolate the effects from the main operating syste,. The
AI "brain" can be any of the usual LLM models from companies like
Anthropic, OpenAI or Google, which run "in the cloud" (ie on one of
their big data centres) but could also be one of the "mini LLMs" like
OLama from Meta (aka Facebook) that actually run in a container on
your own hardware.
Hope that helps.
On Thu, Apr 16, 2026 at 02:40:26PM -0600, Gillian Densmore wrote:
I just found out about podman and pods. Alas the specifi language used
(containers, systemd, system.init etc etc) goes over my head: but I think I get
the gist which is where I want to blow this wide open to yeet out a
litterally complex programing question:
First can somekind person Eli5 to someone who is...kind of a retard: Containers
keep programs (and files permissions) organized (ish)- right?
Without starting a holy war:
Eli5: what on earth the first parts are: systemd: System: Demon but-? I get soo
lost aft that. Then the mind bender:
What kind of black magic is it that I can (try to) get cuttedge tech: ai, in a
sandbox running the background. And yet: when just this morning got ideas
because I was kind of a retard and accidently suggested it was fine to try to
delete my boot drive. I said see if you can figure out why my Linux upgrade
refuses to stay the boot disk. I didn't think that through that She might
attempt remove things to make it organized. the trip:
Wowza! that is a living(ish) thing and me miscomunicating it's like: wow we
have Ai(ish-very ish) So how was it that without to much understanding of
podman: despite setting it's walls as read only (read only permissions if i set
things up right) my little Ai buddy(sensabily) tried to escape rather than
making a text file like I asked. Oops well COOL because that means me and this
agent LLM need to work on comunication skills
But then: wow: mind, blown. If a few poeple would like to toss in a deep dive
to how containers work(complex apps--litterally complicated): so I can
understand this magic better, and also soak in youtube so I don't
acidintally spam all the tech that I just don't understand.
Ajacent to that: to me, this makes way more sense: keeping programs in a
sandbox: just so the rest of the computer doesn't become a mess and also since
I do want to have...agencts? ai? what's the terminogy for something that sits
in the back to randomly do things and I interface with text and profiles? or
is that just LLM? I want to start right on my learning ^_^ danke.
And once again Steve Mea culpa on the excitement about going to space and
generally really neet to see humans getting to gether but still: Mea Culpa.
.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ...
--- -- . / .- .-. . / ..- ... . ..-. ..- .-..
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom
https://bit.ly/virtualfriam
to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ...
--- -- . / .- .-. . / ..- ... . ..-. ..- .-..
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom
https://bit.ly/virtualfriam
to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
