AmigaOS 3.1.4 successor wishlist

[Gorf]

Originally Posted by lantus360

open source and put it on github.

This!

A way to have the amiga undo filesystem actions I did.
Think of dynamic snapshots. You have a system that works, you turn on the snapshot function. Then you change something and the system doesn't work like it should or not at all. Turn off, the dynamic snapshot and roll-back to the moment it was still functional.

It would require the OS to backup all files, that its going to open for write and store it somewhere else, but since the 4GB partition limit doesn't exist anymore, storage is not the limiting factor. Restore is simply a copy back to last know good state.

[demolition]

Originally Posted by Syntrax

Think of dynamic snapshots. You have a system that works, you turn on the snapshot function. Then you change something and the system doesn't work like it should or not at all. Turn off, the dynamic snapshot and roll-back to the moment it was still functional.

Something like the ZFS filesystem ported to 68k?

Originally Posted by Gorf

This!

Making it open source has no value! Directory Opus was made open source and...... were still waiting. Open source is nice if there are 1000's of coders willing to contribute. There are not that many coders left, that can and want to contribute to the amiga. It seems bounties also do not work. MorphOS has many bounties, still waiting for a programmer.

I'd rather see a company pay a programmer to deliver. Then I pay the company. Just my thoughts

Originally Posted by demolition

Something like the ZFS filesystem ported to 68k?

Doesn't need to be in the filesystem Introducing that in the filesystem introduces a lot of complexity. If Snoopdos has a function that looks at the files that are being opened real-time, I suppose there is a way to halt: copy file, continue operation. I've had enough situations where I thought I had only changed one file, but that cascaded in to a chain of other unwanted changes that I had no idea of. So when I undid that change, the damage was already done.

[Gorf]

"btrfs" on Linux or "zfs" on Solaris offer this as a built in function.
And there is of course the Hammer FS from Matt Dillon on his DragonFly BSD.
(Yes it's the same Matt Dillon that did the Dice C-compiler for Amiga)

And there are "OverlayFS" and "aufs" that provide snapshots on top of other filesystems on Linux.

The latter approach could probably be done on Amiga the same way MuFS provides ownership or others that provide compression.

[Daedalus]

Storage space and the 4GB problem haven't been an issue for decades.

Yep, it's relatively easy to hook into file operations like that. The real issue I see with that proposal is speed. Copying a file every time you write to it will at the very least double the time it takes to save anything. On devices where the write speed might be less than 2MB/s, that's a significant addition. Editing the tags in an MP3 could go from less than a second to 10 seconds per edit for example, during which time the application will be frozen because it's waiting on the write process to complete.

It would be a nice feature to have as an option I guess for people who are ok with it. A more traditional backup tool included with the OS might be nicer - I've used BackUp for that purpose for many years.

[Gorf]

Originally Posted by Syntrax

Making it open source has no value! Directory Opus was made open source and...... were still waiting.

If was done a little bit too late in the game ...
But what exactly are you waiting for? It works on AROS.

Open source is nice if there are 1000's of coders willing to contribute. There are not that many coders left, that can and want to contribute to the amiga.

Because very many left and use Open Source software now.
I hear this argument again and again over the last 20 years, but fact is also, that closed source did not help at all to keep developers.
So what about we finally try a different approach instead of repeating the same mistake?

It seems bounties also do not work. MorphOS has many bounties, still waiting for a programmer.

MorphOS is not OSS.
On AROS many bouties worked very well.
Sure others did not, but that can be said for many closed source projects as well.

I'd rather see a company pay a programmer to deliver. Then I pay the company. Just my thoughts

Or you pay a company and still don't get promised drivers or multicore support like on OS4 ... while AROS has SMP now.

Or the company does not pay developers at all und keeps all your money and the source closed, like with 3.1.4


Again: there is nothing wrong with paying developers - the opposite.
Many Linux and BSD developers are paid for their work.

[gulliver]

We talked with Olaf about this sometime ago in a forum.

I asked about a logging filesystem. As it would be very good for version control, roll backs, snapshots, and flash media support.

He was more inclined towards a journaling one due to less complexity, more performance and less wasted space.

The problem I see with ZFS and other related filesystems is that its overhead is prohibitive for Amigas.

Whatever filesystem gets implemented for AmigaOS, it must first and most importantly consider its hardware peculairities.

And then the sad reality comes to bite us: we have too many filesystems for a rather small user base and then we have even fewer developers, which makes priorities shift somewhere else.

[pipper]

For me one thing shines about open source: software preservation. With closed source there’s always the danger of companies going out of business, taking the sources down with them. No chance of getting anything improved or fixed.

[Gorf]

Originally Posted by pipper

For me one thing shines about open source: software preservation. With closed source there’s always the danger of companies going out of business, taking the sources down with them. No chance of getting anything improved or fixed.

It happened quite many times that even popular software on the Amiga was simply abandoned like e.g. IBrowse and the owners/developers were not even interested in selling keyfiles anymore ... but also refused to release at least the keyfiles let alone open the source

Not only functionality is lost this way but also a great deal of knowledge how things can be implemented.

Sadly the Amiga-scene had a habit of doing things in secret and not letting others participate ... in some cases to this very day.

[kolla]

Originally Posted by Syntrax

Making it open source has no value! Directory Opus was made open source and...... were still waiting.

What are you waiting for?

[kolla]

The filesystem issue is just one more thing that makes running AmigaOS hosted ontop of a real operating system a much more attractive solution.

[Bruce Abbott]

Originally Posted by Syntrax

Making it open source has no value! Directory Opus was made open source and...... were still waiting.

If you want to change or add something, do it yourself! That's what 'open source' is all about.

Maybe nobody has done anything with Directory Opus because they are happy enough with how it works now, or perhaps (like me) they would rather code their own applications than try to put lipstick on a pig.

I am not against open source in general, only the unrestricted open-sourcing of an OS that might encourage a plethora of incompatible versions.

[Gorf]

Originally Posted by Bruce Abbott

I am not against open source in general, only the unrestricted open-sourcing of an OS that might encourage a plethora of incompatible versions.

I do understand your concerns.
But actually that is NOT what is happening in real life open source software.

We still got only one Linux kernel and despite of very different distributions, a binary (even closed source) can be compiled in a way, that will allow execution in every one them as long as it is the same cpu architecture.

We got four mayor BSD flavors, but still they provide a compatible API. Even more: you can run the Linux-binary from above on them!

Now compare that to our closed source OS4 and MOS

Yes: there might be a "a plethora of incompatible versions" on GitHub or on hackers personal computers, but Darvin's law is hard - only one will prevail.

[Marle]

Of course looking back to 1993 is probably not the best way to go forwards, but Commodore's plans for AmigaOS presented at the Orlando Dev'con' in 1993 are interesting to read:

http://obligement.free.fr/files/devcon93.pdf

Might be a few lower level things in there that would still make some sense!

[Bruce Abbott]

Originally Posted by Gorf

... but also refused to release at least the keyfiles let alone open the source

Aweb was open-sourced, and man what a mess! I wouldn't know where to begin trying to fix it.

The problem with open-sourcing a large application is that just getting it to compile can be a struggle, let alone understanding its inner workings sufficiently to make changes with confidence. And when it has fundamental flaws that require a complete rewrite...

IBrowse 2.5 was still in active development as of 02-Nov-18, so I am hopeful that it will be released soon. Would open-sourcing have helped? That depends on what the code looks like, and whether anyone would be bothered spending the time to understand it.

Originally Posted by Gorf

We still got only one Linux kernel... We got four mayor BSD flavors, but still they provide a compatible API.

You need a lot more than just the kernal to run most Linux apps.

list of Linux distribution problems

Ubuntu developers decided to push Ubuntu as a viable gaming platform and they identified the topics which need to be addressed in order to achieve this goal. Uncannily the list, they've come up with, matches the list, you can read below, almost verbatim. In 2017 Ubuntu (as most other Linux distros) still struggled with GPUs, HiDPI, Network Manager and two dozens of other problematic areas.

Originally Posted by Gorf

Now compare that to our closed source OS4 and MOS

Not having either OS I don't know much about them, so I rely on second-hand information. Wikipedia says:-

MorphOS

Developed for PowerPC processors from Freescale and IBM while supporting the original AmigaOS MC680x0 applications via proprietary task-based emulation, and most AmigaOS/PPC applications via API wrappers. It is API compatible with AmigaOS 3.1

OS4 kernal Wiki says:-

Exec was originally written in 68K assembly code. The migration to PowerPC-based systems necessitated a rewrite of Exec while retaining as much of the original API as possible.

[gulliver]

Originally Posted by Marle

Of course looking back to 1993 is probably not the best way to go forwards, but Commodore's plans for AmigaOS presented at the Orlando Dev'con' in 1993 are interesting to read:

http://obligement.free.fr/files/devcon93.pdf

Might be a few lower level things in there that would still make some sense!

An interesting document.

Many things have been already incorporated into 3.1.4.

Some nice mouse related ideas.

Others, specially the gadget extensions are something we all probably want and need, in some other form like ReAction or MUI.

He also suggests to add consistency by regrouping functionality into single entities. Something I have allways pushed forward, but seem to find no supporters.

He only briefly mentions Workbench, which is what I believe should need most rework.

Thank you for the read.

[Gorf]

Originally Posted by Bruce Abbott

Aweb was open-sourced, and man what a mess! I wouldn't know where to begin trying to fix it.

OK i've got one for you: IBrowse.
No idea it it is a mess, nobody will ever know. Quite some people would like to buy the necessary keyfile, but nobody is even willing g to sell it anymore - let alone publish it or - "oh no!" give the source free.
Hail to closed source!

The problem with open-sourcing a large application is that just getting it to compile can be a struggle,

sure - that is why you need to do it in an orderly manner, documenting the whole build process and so on.

let alone understanding its inner workings sufficiently to make changes with confidence.

that is why you always need sufficient comments in the code and a good documentation.

And when it has fundamental flaws that require a complete rewrite...

sure - that is true for all projects, open or closed.

IBrowse 2.5 was still in active development as of 02-Nov-18, so I am hopeful that it will be released soon.

Oh the situation has changes there?
Not so much according to Amiga.org ... the thread is still discussing if hacking IBrowse is a possible solution ...

Would open-sourcing have helped? That depends on what the code looks like, and whether anyone would be bothered spending the time to understand it.

Again the same mistake:
no it does not really help to open the source at some later point.
But developing it in the open it would probably have helped to maintain a readable und understandable code.
(No not "automagically", but bad code and undocumented projects simply won't survive long)

You need a lot more than just the kernal to run most Linux apps.

Not really - all my docker containers need is a kernel. Batteries are included.

Not having either OS I don't know much about them, so I rely on second-hand information. Wikipedia says:-

Wow both can run legacy Apps!
And so does WinUAE

Sorry Bruce, but that is ... You know what.

[Gorf]

Originally Posted by Marle

Of course looking back to 1993 is probably not the best way to go forwards, but Commodore's plans for AmigaOS presented at the Orlando Dev'con' in 1993 are interesting to read:

http://obligement.free.fr/files/devcon93.pdf

Might be a few lower level things in there that would still make some sense!

Thank you!
Very good read!