I agree with you that "Packaging system was waste of time”, but people do what they want, it is voluntary project.
I think Haiku pack.system mimics MacOS package system. Yes, it looks cool and funny maybe modern to some people, but I think it is too complicated.
I liked more GoboLinux system when every app and lib have it’s own directory and this is enough for differiantiating versions of software, file system by itself do the job.
Maybe some one want to make spinoff of todays Haku? And make system as they like.
If Haiku not make to beta soon, I think this will happen.
Personally I liked more system with different name (maybe OBOS?), GoboLinux like app and lib managing system and not to stick to MIT license. I think is better policy to give the way to best software not to some license. People use software not licenses.
I agree with you that "Packaging system was waste of time”, but people do what they want, it is voluntary project.
“too complicated” is surely no argument to the end user, he sees applications popping up in the deskbar applications menu.
Oh yes, but for that we already have Ubuntu and etc, MacOS, Android, Windows.
But some people (I think) and I like BeOS way more: simple, efficient, elegant and easy tweakable.
It depends on what you use the package manager for (or what goal it is designed for in the first place). If it’s more about “I want to find cool stuff I can install on my Haiku” then it’s no problem to have a package manager. And update nowadays are quite frequent so having the latest version delivered without having to hunt it down is not that wrong. Definitely easier for the end user than trying to install from websites (I did this in BeOS days… and it has been not funny sometimes).
Does library hell rings a Bell for you?
Thanks to the PM we already ported and updated programs was without PM would ve just painful.
The dews doesnt need to use PM, they can just Put everything in non packaged.
But why Do we have Talk it again and again?
I am not against Package Management, I am against that how it is done in Haiku: it is complicated solution – it made system unnecessary more complex.
Problems with library versions can be resolved in more simple way: putting those version in different directories or/and with naming it different, no need for some virtual FS and additional special things in system. And you still can have some PM app to install/uninstall software – it is separate thing.
You are free to create packages out of haiku ports or git. For open source project a nice way to share. Closed source projects would be use the command line or other package creator to make hpkg files. You can generate youe own repository server too. So many things are possible. The package Management system of haiku is very great, i Like it. Ok, you can make things every time better like original, but this most times better for you and not for all. If packages are created intelligent you have the possebility to install all dependencies too, out of one installation process. In old times you need to do this many times by hand.
You changing subject. You see?
I am talking about system structure efficiency, not about user’s experience with app.
Trying to understand: you like how GoboLinux unpacks everything on the filesystem, and doesn’t like how Haiku doesn’t unpack everything on the filesystem, right?
@damoklas, you know, what we call PM today is not the final form of the proposed package management system, just a development milestone, the development frozed to let the beta happen, after that it will be polished further to let it shine.
Critizing the current state is absolutely okay, but you know well we had tons and tons of comment like yours, and you know well, what was the answer:
“If you don’t like it, then please create a ticket at the bugtracker with your ideas how the devs can make it better, semless, integrated, easy, good, and nice. This is the preferred way btw.
Or as alternative fork the whole stuff and make it PM-less. To help you, here is the latest commit before the PM-merge: 23afdc2ba493acf05dd9bd892e95c568d0d35670.”
We absolutely talked this over and over, so please do not expect any big changes before the beta, nobody will remove the PM subsystem, as it works well and the developers know really good, why it is made and why does it required.
We can understand if you are frustrated, but writing comments like this won’t change anything. The PM is merged, time lost, okay, happend, bad, but please, try to get over it and take part in the developement with ideas, translations, ports, or art.
That’s the best way to make beta happen and to make it shine.
No, not right. You simplify too much my position. I said:
And I can add, that with New Memory (RAM+ROM or even SDD) technology current Haiku package-VirtualFS solution are/will become overdo and obsolete.
Thanks, I investigate this possibility. Unfortunately this is above my skills…
But seems this is only way to get what I want.
Only question now: want I enough for doing that by my self?
Only you can answer that.
“booming”? A few years ago, everyone went something like “the OS is great, I would use it if only there were apps for it.” And apps wouldn’t emerge because they would always be broken in some way or another, and needed endless tweaks to libraries to keep things more or less running.
I spent countless hours messing with software that would try to get me to copy libs to system folder and would make my system unbootable in very creative ways. With package management, this problem is solved, there are much less ways for user to break their system, and even if they manage to, it’s possible to boot an older version/state anyway.
This is starting to work, as you can see with many KDE applications now running on Haiku (and that is only the latest addition, we have more and more ported things every day).
The package management system has a great “wow” factor when demonstrating it to other people as well. Why do you think we lost developers? I’d say it’s more likely that we gained some.
I would probably not be using (and then, contributing to) Haiku if I had to spend days and nights re-compiling my tools to get things done. Thanks to haikuporter, I can now have these tools in the repos, and just need to make small fixes and updates to recipes once in a while. This means working with Haiku is now faster for me, because on Linux, none of these tools are packaged.
Anyway, if you think package management is a waste of time, fork the code from before it was merged, get in touch with the developers we lost, and assemble a team to show us how wrong we are. We’ll see what happens
Looking back my comment was probably a bit too strong – I am a big fan of how Windows handles applications, but because many old beos/linux apps don’t have good installers I do see your point. I will wait until beta before I pass more judgement.
Thanks for the kind response - I probably didn’t deserve it
Wouldn’t it be great if there was simply one way of doing things (like it was with BeOS)? Because this is a free community effort (not a closed-source corporate one), it isn’t that simple because EVERYONE has their own ideas for how to do things, and if something CAN be changed, it most likely WILL be changed. And again. And yet again. This added and that changed and these moved and those removed. Etc.
Things that worked before, break. Users complain. Devs push back with “Its Alpha, expect it.” The cycle goes on and on. It’s been, now, 17 yrs.(?!?) and we STILL haven’t reached beta! Did it take Be, Inc 17 yrs. to release BeOS DR8 (which probably equates to their version of beta)?
But they had money to bring their singular goal to fruition. Insufficient as it may have been, it was pretty much “final” when BeOS R3 was released, as I recall. It was what it was and we liked it. Then came all the other revisions til R5.0.1. But even if we assume the very most generous premise that “Alpha” was DR1 and R5 was “final”, that only took a total of 5 yrs.!
One man lead the charge… Jean-Louis-Gassee’. And everyone followed his direction. THAT is how you get an OS built in 5 years. Well, that and sufficient funds to execute that direction.
Haiku doesn’t have that kind of focused attention anymore. It once did, and that is what got Haiku going like wildfire in the beginning. Passion and dedication to a goal is what will get you farthest.
But what IS Haiku’s goal nowadays? R1 = R5? But Haiku already IS better and more advanced than R5. You can do more with it, faster, than R5 ever could, can’t you?
So, then… where is the goal now? WHAT is the goal? And I think that is why it’s taken 17 yrs. to get to this “almost to beta” point. Everyone is still “adding this” and “changing that” and “moving these” and “removing those”.
You HAVE to have a specific goal to reach. No wandering off on bunny trails. You have to find the people willing and able to accomplish the tasks that need doing and do whatever you must, to insure they COMPLETE those tasks. No ifs, ands, or buts about it.
I would (if I could) see a version of Haiku that was so secure, NOTHING and NO ONE could break through. THAT is my first and primary goal. But the way I seek to achieve that goal is so radical, it’s NEVER been done. Not in Windows. Not in Linux. Not in MacOS X. Why? Because it cannot be retro-fitted in. It must be built-in, from the ground up. And, it would BREAK the way EVERY OS made today operates. Seriously.
But radical departures from the status quo are what made BeOS and MacOS X.
Haiku needs that kind of focus and purpose once more…
Hmmm… Be. Inc was a company and haiku is a open source project started by a group of friends to get the ghost of beos alive. Thats all.
I am not a christian, I am not interested in that. Your wrongness is all yours.
And I suggest we stick to the matter.
Even though I’m not crazy about all that Linux-style dependency world and locked-off system with read-only folders (I see why it might be especially frustrating for devs in comparison to pre-PM Haiku times), I like the idea of cleaner package system. I’m more into Mac-style (all-in-one) app packages, they are huge, but pretty much self-contained. Also, it is true, that HDD space and internet bandwidth is not a concern now, very different world from 15+ years ago. Of course, ideally we want lean, mean and crazy fast native apps in tiny packages, but for now all we have is old BeOS apps and Qt/Java ports (via HaikuPorter or not), which bring lots and lots of dependencies with them. Saying that, online repos and self-updating mechanism is a huge step forward for Haiku.
Pre-PM Haiku was like a big ol’ sandbox. Now it’s like Android with even more strict rules - very easy to install and update apps, but almost impossible to mess around. For better or worse.
There is the directory /home/config/non-packaged. In that directory you (and your users) can do absolutely anything you like, just as it was in BeOS. You can use Haiku as your base, ignore the package system and develop your own ideas in non-packaged. You can develop your own packaging system that simply sticks to non-packaged. You can symlink it to /DioGenOS if you like. That way you can update the OS from the official repos, and present apps to your users in your own way. Nobody is stopping you.
Try it now. download a user app hpkg with not too many dependencies from the depot web app, Use Expander to extract the contents to non-packaged The app will appear in your deskbar and you can fiddle with it as much as you like.
Right now, yab developers who want the latest bleeding-edge version of yab use a script by bbjimmy to clone yab from github and compile it. It bypasses the packaging system. So it can be done. Go ahead and impress us.