Quoting my user profile:

I like fixing turing-mechanical devices (big banks of switches like LCD displays and microprocessors etc.). I am a 4th generation doctoral descendant of Alan M. Turing, and the only one I am aware of in England south of Bristol and West of Oxford. I have been into GNU and Linux and X-Windows since my first year of uni (1995-6) where I did a year of physics before switching to maths and getting a Ph.D. under Richard Kaye, who wrote a certain paper on how hard minesweeper is for a machine to check.

I am interested in taking Antix as it is as a basis for my future custom Debian derived distros. I will use Ubuntu Studio for a collection of working creative software to get started. I am happy to learn etc.

The ghost of A.M.Turing says hello, BTW.

John

(google: John Allsup mathematical genealogy and go 4 hops of the supervisor chain to find Dear Alan. I am in the process of trying to found the UK Guild of Turing Mechanics and the UK Association of Lambda Theorists, and have registered the webspace. It would be nice to cooperate with you guys.)

Check out
========= SCRAPER REMOVED AN EMBEDDED LINK HERE ===========
url was:"http://john.allsup.co/makepass.html"
linktext was:"http://john.allsup.co/makepass.html"
====================================
for a hint on password management.

Welcome and Howdy! Its nice to see you want to learn some more.

AntiX was made to help older computers to stay active. Lots of newer distros leave the older machines in the dirt, unable to run the OS at all.

wondering:
Was the now abandoned turingos.org (and the associated distrolette) your endeavor?
What was your 2013 epiphany?
Which"creative software" titles
========= SCRAPER REMOVED AN EMBEDDED LINK HERE ===========
url was:"https://ubuntustudio.org/tour/"
linktext was:"https://ubuntustudio.org/tour/"
====================================
preinstalled by Ubuntu Studio do you find appealing?

it's not openGL, but the code is available on github and you might find this to be interesting:

========= SCRAPER REMOVED AN EMBEDDED LINK HERE ===========
url was:"http://geokone.net/"
linktext was:"http://geokone.net/"
====================================

Re turingos: no.

My OS project, as yet unpublished, is KOS. Killer Operating System (from Heaven/Hell as appropriate).
The aim is as follows.
Use Linux Kernel.
Write lightweight forthish layer (Glish is my concept: 64bit word based with no string processing, since
things like 'docjohna' and 'mrflibbl' fit in 64bit words, and we can use byte ordering so that lex order
agrees with dictionary order, including defining a new character encoding, I call jamz notation, so that
we have 0=NUL,1=SPACE,2=A,3=a,4=B,5=b,... and then we can do basic lex ordering with no string
processing provided 'strings' are less than 8 chars -- basically work base 2^64 and emulate what forth
did on 8bit processors using 64bit bytes, leaving the interface to more conventional string stuff to a higher
layer.) Then LLVM and an scripting language (pythonic, I call it Sidewinder) to script what gets compiled
to binary. Importantly there is no direct facility for loading binaries, so all code must be LLVM bytecode
or script source, and must include license details in metadata, and software license will make this binding.
This is why RMS is wrong about LLVM: we can do Free Software better using what LLVM provides, and
outcompete proprietary solutions if we get stuff right.

I am a naturally gifted machine computation theorist, who did a Ph.D. under Richard Kaye. My thesis is number
20 at john.allsup.co, and note the note on page 7: I am a software artists, not an engineer. I do not generally release
anything that I do not consider beautiful (like TeX and TAOCP, but not LaTeX or your average C++ program).

I would like to hack PhpBB, simplifying it. If you look for Wabisabi, the wiki, you will find a barebones textfile
based traditional wiki. This needs a bit of organising, and then we can have these all over the site with individual
passwords and filestores, and do away with the need for an SQL database for anything except indexing and storing crypto hashes to check integrity. That is what MySQL and SQLite are great for, but if you want a filing system, use the filesystem.
Recall what Chuck Moore of forth fame said: the filesystem is a database, and for your application design, get the database right first.

I intend to resurrect old school thinking re lisp and forth and show how to make things fly on modern equipment. I would like to take a source tree of the current AntiX, get it building on my HP Z800 (with 12 cores, 48GB ram, can build Linux kernel from scratch in 8mins and most things in a few seconds using make -j24). UK based residents are advised to contact me with regards to bang for buck on a budget advice, since I had to spend years on disability benefits rebuilding my brain, buying cheap and maintaining it myself (including using workstations and servers as fan heaters to just take n*100W out of the wall and dump it into a room as heat, which they do better than radiators since there is no piping to leak heat __{{emoticon}}__ )

My personal website is john.allsup.co, I have registered turingmechanics.{uk,org} and lambdatheory.org and lambdatheorists.uk so as to start the work of building the language humans use to talk about modern computation along the lines Turing would have developed it. Do recall the publicly available video by the authors of StructureAndInterpretationOfComputerPrograms, when one of them said that 'computer science' is a historical accident: I am a computer and my machines are my peripherals; and science cannot handle computers: only Mathematical Logic can, and I am a Ph.D. qualified Mathematical Logician from the lineage of AMTuring.

Finally, how to I get a modified antix so that there is a, say 256MB ramdisk with all main utilities and basic compiling software in it. LSBase has to go IMHO. My layout should be:

/boo/{kernel,etc} -- boot stuff
/sys/{bin,sbin,lib} etc -- basic system stuff
/usr/jda/{bin,sbin,lib} etc -- jda stuff (each of us has an initialism, mine is jda, and our stuff goes in there)
/usr/rms/stuff
/usr/linus/stuff -- git would go in here

Each programmer domain would have its own hash, and should reside on one filesystem, but /usr/jda and /usr/rms may be located elsewhere, possibly mounted via a web based protocol or sshfs.

Importantly, the above approach minimises load on the machine and, with a little practice, cognitive load on the user.

Finally, with regards to systemd, the concept is good, but implementation suffers from brain failure issues. We need a modular init system (this is what my glish concept is there for) so that for simple systems where a DAG based mode switcher is not needed, we are not forced to have one, and where we do need one, such as a modern laptop, we are not forced to engineer thousand lines of unreadable shell scripts. And I do not want to write a single line of C++ if I can avoid it.

John

John

Howdy and Welcome. __{{emoticon}}__

UK based residents are advised to contact me with regards to bang for buck on a budget advice, since I had to spend years on disability benefits rebuilding my brain, buying cheap and maintaining it myself (including using workstations and servers as fan heaters to just take n*100W out of the wall and dump it into a room as heat, which they do better than radiators since there is no piping to leak heat

Kool. I am a tinkerer myself but self taught and not any certification like you have.
My grinning personality gets me free obsolete city hall servers every once in a while.
My forte is motorcyles rather than computing. But I used to be known for fixing a helicopter or two.

I will not pretend to understand 1/2 of what you are saying as it goes"whoosh" flying over my head.
Sounds cool as all get out though.

Happy trailz, Rok

First, any of you who have not read Thinking Forth by Leo Brodie, either download a retyped copy for free, or else buy one from Amazon. The actual implementation languages are, of course, out of date, but the thinking behind how to use it is timeless and priceless. Do likewise with the Lisp manual 1.5, especially observe what the guy behind Smalltalk said of the eval function around page 15: the Maxwell's equations of software. Modern programmers do not need to use Lisp and Forth, but all programmers should understand the philosophy behind them: a kind of pragmatic ultraminimalism. I aim to demonstrate that with Linux and whatever I deem appropriate to run atop it (sorry GNU, but you're a little on the large side).

I would like to:
1. Produce a personal fork of the current version of antiX, and then pick it apart to understand it.
2. Retrace the development history of antiX (hopefully some of the older developers are still on this forum).
3. Develop some ultrasimple web software, using python3 for offline jobs, an ssh backdoor for admin authentication, minimal use of Php (i.e. not a 250,000 line behemoth, but rather something around 512 lines of Php all in, like WabiSabi is for a wiki). For me function comes first and all else is optional, and cutting cruft is what I was trained to do to mathematical calculations anyway, so it will be fun to dust the rust off my university brain.
I need to:
1. Know how to acquire the source tree and knowledge of how to do a complete build. I have a 12-core Xeon workstation to do the heavy lifting, and it has little else to do except boot into Windows and run my Vst collection. It compiles a stock Linux kernel from scratch off a standard spinning disk in about 8mins, if I do a make -j24. That's fast enough for what I want, system wise.
2. Get hold of the build tree of the first 64bit version. (I will be exclusively 64bit for non-embedded stuff, and then produce a cut-down Debian derivative for my RPi collection next, applying the same philosophy once the concepts are worked out.)
3. Get in touch with whoever does your web admin if possible for friendly conversations. If I can produce working code that does stuff better, it is yours under whatever free software license the original is (just acknowledge where the bits came from, which I shall do by appending #JDA, //JDA or inserting /*JDA: ... \nEND JDA*/ as appropriate. Later I intend to develop my programming interests so as to move away from character based source languages. What I think is best is to construct an object (a theoretical model of the working binary) in memory using modern scripting techniques, and then instruct the compilation infrastructure as to how exactly I want it turned into working code. I do not like my computational devices to try and guess for me, since they get stuff wrong more often than the get it right. (For me, -O3 is like a Clippy boucning around going 'It looks like you're trying to refactor your code, let me autobodge it for you so that it goes twice as fast and only crashes occasionally' or something like that. Modern compiler optimisation is basically a game of writing an AI to guess how the programmer should have written their low level implementation, given essentially no actual understanding of what the program is meant to do, nor how fast it is intended to run. See Knuth's comments on optimisation and the need for discipline in its usage.)

Anybody interested in an informal research arm of the antiX project (I am happy to discuss and prod etc.)

John

I need to:
1. Know how to acquire the source tree and knowledge of how to do a complete build.
2. Get hold of the build tree of the first 64bit version.

Except for a few in-house scripts/utilities, which are published to

========= SCRAPER REMOVED AN EMBEDDED LINK HERE ===========
url was:"https://github.com/antiX-Linux"
linktext was:"https://github.com/antiX-Linux"
====================================
(possibly a few still reside in other repos, at the moment e.g. gfxboot, in
========= SCRAPER REMOVED AN EMBEDDED LINK HERE ===========
url was:"https://github.com/BitJam"
linktext was:"https://github.com/BitJam"
====================================
)
the"source tree" is Debian. I don't know (and don't care to know) how one would retrieve the entire source tree but,
for individual packages of interest, I visit
========= SCRAPER REMOVED AN EMBEDDED LINK HERE ===========
url was:"https://www.debian.org/distrib/packages#search_packages"
linktext was:"https://www.debian.org/distrib/packages#search_packages"
====================================
and use the link provided
on the package's detail page to download a source tarball for the package.
antix does ship custom kernels & I'm not familiar with the source/build details; someone else can chime in to provide that info.

As for the rest, (moving away from character based source languages)(dafuq?) ...good luck with that.

edited to add:

how to I get a modified antix so that there is a, say 256MB ramdisk with all main utilities and basic compiling software in it.

By choosing antix as the base, you would be starting with a unnecessarily"bloated" (for your needs) kernel, chock-full of backported drivers.
You would be better served by building ala Linux From Scratch
========= SCRAPER REMOVED AN EMBEDDED LINK HERE ===========
url was:"http://www.linuxfromscratch.org/"
linktext was:"http://www.linuxfromscratch.org/"
====================================

For now, I will stick my own thougts on this thread rather than launching multiple threads on the board. It makes searching much easier since, unless I am replying, my thoughts are all in once place. When this thread gets too long, I will launch a new thread.

From my point of view, a computing machine is just a bank of switches, and the software (whether firmware, kernel, compiler, etc) is just a configuration of those switches. In practice, there is one big global computing machine (a heterogeneous computing device whose interconnect is the internet). That is how I am trained to see such things, and it makes many things much simpler. You do need a lot of mathematical training to handle that number of switches, but that is what I was trained in (maths with numbers provably bigger than you can count up to).

Now,

Another thing I am interested in helping develop is a custom install maker. The use cases are:
1. Write a script, give a target block device and place a working image on that block device.
2. Generate a special purpose image. For example, if I want a USB stick for digital painting, have an image with the software for that, plust the basic command line stuff (I would include a minmal XeTeX environment and a large collection of fonts too). You would end up with something less than 256MB in size, which could be stored as a small number of image files on a USB key, and loaded and run out of RAM. (So instead of a 4GB laptop, you have a 3.5GB laptop with all system stuff on storage faster than the fastest SSD. Essentially, the ideal place to put this, following the example of the RPi, is on an SD card, and for security make it read only and use a USB key for storage. For storage you have a versioned store of some sort which is another 256MB RAM for all config data and small files (word processing, text, source code and spreadsheets) and occasionally commit this to the memory key. Larger media is then handled via a USB3 SSD using btrfs when it is ready, backing this onto a RAID array of spinning disks (which have SMART and fail more gracefully). Importantly, the hardware heavy lifting can be done on the inexpensive ex-corporate refurb servers you pick up on ebay. Get reliability through redundancy, and if you stick the server rack at the bottom of your house and use the exhaust as auxiliary heating, you can save the cost of the electricity back from your central heating bill. In summer recess you power it all down for annual maintenance, take a holiday, and do not do any mission critical work, using hardware overheat issues as your excuse. B-)

So, a special purpose image generator is something I want to help develop. What you want is something which runs under standard Debian and dumps the lightweight image straight onto the target disk. Step one would be to have a script for current Debian which essentially reproduces the current antiX on the target block device.

Any thoughts?

Since this is a tech people's forum, I thought I would put this here for now. At some time I will look for a better forum, but finding the 'only publish with us' or 'patent it and milk it' attitudes distasteful, I would rather just stick good ideas out there so that people can read them, think them through, and try to implement them. As Kepler said: My witness is the Bright Daylight and Time.

Now,

Modern Ideographic Language
---------------------------

Have a standard spec of the shape of e.g. shopping trolley, or play.
Include latin alphabet in this too. Do likewise for accent shapes.
Have a means of programmatically combining subglyphs, and have a
64+bit space for glyphs. Essentially a glyphstream looks like forth.

context :name :url set
capital latin-a latin-l twice latin-s latin-u latin-p stdpunc. latin-c latin-o
lang English esv set
capital latin-i latin-n stdspace latin-t latin-h latin-e stdspace etc.

The glyphgen then takes this and sets it. Either a penstroke method
like metafont or an as-yet-unwritten modern digital calligraphy system,
or else an outline font system such as opentype, or something based
on SVG. Untimately this frees us from the straitjacket of unicode, which
is in the no-mans-land between ASCII and full expressibility.

One could take the wire-protocol idea and apply it here, but it would
be better to have a dictionary of standard use cases and use an appropriate
encoding scheme with cryptographic hashes (sha256) for error detection.
Essentially you have 256bit blocks of text, and hashes at predetermined
points. The hash starts with all zeros, and takes the stream up to a hashpoint.
Then we hash, check it with the intended hash, and if ok, use this hash
as the first block of the next chunk. All that is needed is a list of
blockcounts and hashes. A sensible system is to have one hash per 4k block,
chaining as above, if we want a package. If we want a jigsaw, we just
hash each block separately and concatenate all the hashes and hash that.
One can encrypt with aes256 first (keep block size compatible) and then
do this. By jigsawing and and package hashing (so use two hashes per 4k)
we can jigsaw to get the pieces, and find which block takes a zero inithash
and checks out, and then proceed searching the pieces to find the correct order.
If we have correct pieces, they will fit.

I call this the cryptographic jigsaw technique.

want is something which runs under standard Debian and dumps the lightweight image straight onto the target disk.
Step one would be to have a script for current Debian which essentially reproduces the current antiX on the target block device.

Any thoughts?

Obviously, somebodee did not RTFM.
keywords for your search: persistence, frugal, cheeseburger, snapshot, remaster, isosnapshot, antix2usb

RTFM is so last century. These days TFMs are TL;DR. If I need to RTFM at all, that rules out even trying. If you want to constructively suggest RTFMing, please either post a tested link to a google search or a link to the appropriate online manual. 'You do the work because I do not want to' is a bad habit that needs to die.

As I stated earlier in the thread (and please read what I write in order and carefully without guessing) I have a number of thoughts regarding antiX, Debian and progress. I can write it here and hope for constructive feedback with little effort. Since the stream of ideas is confined to one thread, there is little inconvenience for others. But if readers get what I am chasing after, that would be a good thing.

I cannot deal with vague suggestions, having drilled my mind to demand near certainty. Where I wish to go with lightweight Linux is not sufficiently catered for. If anybody can point to a short legible working script which does what I asked, that is the aim. Producing a custom usb stick should be easy, not just possible for a select few. The final difficulty is locating the correct advice within a large haystack.

If you have a particular search in mind, just post links to where the search points. Vague hinting and 'sombodee' stuff is impolite. I have searched for what I am after, and those searches failed. Perhaps what is obvious to you is not what I am communicating. That is why experts in logic are trained not to guess at probable meanings: at the level one is working with when hacking the foundations of a modern Linux OS, probably and obviously are enemies, not friends. Precise advice which is readily checkable is key, and all else is line noise.

What I was trying to do above is describe the vision I am seeking, and get it in one place.

If I need to RTFM at all, that rules out even trying.

At this juncture, perhaps I should just sit back and let Darwinism run its course.

want is something which runs under standard Debian and dumps the lightweight image straight onto the target disk.
Step one would be to have a script for current Debian which essentially reproduces the current antiX on the target block device.

Producing a custom usb stick should be easy, not just possible for a select few.

I'm saying"Good news! What you've described (as 'Step1') is possible and is easily accomplished".
Seems like you're staring into the corner, whining"it's dark in here... and the voices are impolite."

antix15 full version.
From within a liveUSB session, after adding/removing programs and performing other personalized customizations,
you can launch"ControlCentre" via the desktop menu, then within ControlCenter, click"Live" navtab, click"remaster", supply root password...
and the"antiX Live Remaster" tool (script) will walk you through the process.

For any detail/term/phrase which isn't obvious to you during the process, search the forum for relevant discussions or just make your best guess and proceed.

If you have a particular search in mind, just post links to where the search points.

You're missing the friggin' point, Chief.
If"RTFM is so last century"... what replaced it?
Self empowerment via self-directed learning, acquiring relevant knowledge just-in-time, on-demand.