GUI installer

You quip back with generic "Nobody is stopping you or anybody else from doing the work." style comments to somehow prove a point.

It does prove a point. It proves, in the most fundamental way possible, that no-one cares THAT much that they wish to invest the time/money to create/have created for them the [you name it - it doesn't just apply to a new FreeBSD installer].

I find the current sysinstall usable for its purpose both before and after an installation. I'm not alone :) I therefore do not wish to waste my limited time/energy/resources on creating a new one.

I also agree completely with the previous viewpoints that:

1) FreeBSD does not have to be like a new user's previous operating system (witness how the Linux horde is turning Linux into another operating system so that one won't eventually be able to tell the difference. If I want to run another operating system, I will run it - I don't need a quasi-compatible clone);

2) Anyone who is unwilling to experiment/persevere with the installation of FreeBSD using the current installer, is probably NOT going to want to learn enough to actually be able to use the operating system successfully. (This isn't a criticism; some folks don't want to learn how their car works either, they just want to "use it" in which case this car/operating system is not for them.)

And perhaps my experience is also coloured by the fact that the Solaris graphical installer never worked, and the text installer was the only way to generally get the os installed assuming you had the required supported legacy hardware or were really prepared to hack a baffling array of configuration files just to get a network running. FreeBSD is a dream by comparison :)

And finally, yes, I started programming with the command line on a Commodore Vic-20 using CBM BASIC and 6502 machine code. Ah, those were the days...
 
I'll second this one :)

tangram said:
Why a GUI installer?
...
FreeBSD needs to continue it's focus on reliability and flexibility not needless user friendliness.

...

I' don't mind learning from ground up; it's like building a house ground up. So it's should stay that way - reliability first.
 
I figured this topic would be dead. But people keep chiming in on how it shouldn't change and you learn from sysinstall.

Why is it so hard to absorb information into the brain? No one is saying you can only have one install method and a bulky X-based GUI is the way! I take that back -- sysinstall fanboys are screaming sysinstall only and not allowing for other gui install methods to exist.

Both could exist and satisfy all conditions. I still believe the majority would use sysinstall -- that is, if sysinstall fixes some known problems and adds the ability for install-time configuring not allowed previous.

Comments saying "drop to a shell and do install stuff from there" are only half-truths. While a shell to drop to is excellent and I think needed, having to manually install what an installer should do only shows lack of configuration and and failure at its only function.

Stop pretending it has to be one or the other. I don't think that has ever been the case presented nor the path intended. I repeat that the only ones claiming such a scenario are the ones demanding sysinstall remains as the sole installer.

I think I'd love to see a compilation of "knowledge learned" from sysinstall and forcing users to use it. While I don't think you learn nothing from using it. I am hardly convinced sysinstall grants an immeasurable amount of knowledge that can only be obtained by such an installer and thus provides an immense social baptist-by-fire experience of osmosis learning. I think we exaggerate the usefulness of the few steps required by sysinstall.

Frankly I don't give a hoot on the graphical appearance of the installer. I really just want a fully featured installer with the ability to configure all the necessary options as required for the various installs practiced upon the operating system in use. Sysinstall does not currently meet this criteria but if it did then I have no problem being a promoter of its usefulness.

I often use FreeBSD strictly as a server operating system for headless units. I don't ever use graphical implementations on the box ever. The only time a monitor is plugged into the machine is during install, however, which puts all graphically based criterion specifically on the installer. I'd rather not be forced to run a large X-Windows install operation by an only choice but would prefer a similar representative experience within a text-based installer.

Just one last laughable point. To Quote, "FreeBSD needs to continue it's focus on reliability and flexibility not needless user friendliness."

Besides the laudable attempt to intertwine reliability and flexibility of FreeBSD as an OS to the installer, the point is contradicted by flexibility specifically as that is exactly what sysinstall has proven not to be. That even is part of the core debate about sysinstall or have you neglected to read any discussions previously about this?

If you take anything from my discussion, just remember you don't have to pretend the debate is a one-or-nothing installer. Why is it such a terrible situation to have a choice of installation methods? And why does it seem laudable to force a choice on other users just because you had no choice previous? (Actually the last question was pretty eloquently discussed earlier in this thread and seemed to ring true as per this discussion. I bring up this last question for self reflection before people decide to post rantings about "KEEP IT THE SAME! DOWN WITH CHANGE!")

Good day.
 
I find sysinstall functional and highly reliable. I'm just finishing up an install of 7.1. :)
However, I do believe there is value in having a graphical installation method available as well as sysinstall. FreeBSD may be able to attract a larger user base if it is easier to use.
I'm not saying that FreeBSD should be PC-BSD, but, that it should allow more choices for users. Gentoo now has a graphical installer.
 
I'm going to chime too. We still need a text based installer, but a graphical installer would be a welcome bonus. I asked about a GUI installer at the recent MeetBSD event, but was quickly shot down. That's too bad, because the goal isn't one-way-for-everyone, but choice. Why can't I run "sysinstall -g" to get a graphical version?

But in any case, sysinstall is getting dated. There is definitely a need for a sysinstall-ng, with a clean separation of the backend. Then the front end could be ncurses text based, or graphical GTK or Qt. Or even HTML/Javascript for remote installs.

Blindly asserting that the current sysinstall must live on forever and ever, end of discussion, is silly.
 
The principle of finstall is that you have a backend and a frontend of your choice. So the frontend to finstall can be a GUI, some text based installer, or even a webinterface so that you can install FreeBSD over the network on a headless box.
 
lme@ said:
The principle of finstall is that you have a backend and a frontend of your choice. So the frontend to finstall can be a GUI, some text based installer, or even a webinterface so that you can install FreeBSD over the network on a headless box.
This basic idea is why I'm for finstall replacing sysinstall. Also, there are a few people who claimed sysinstall had a "smaller footprint"-huh? What else are you doing on a box at the same time as installing a new operating system? It's not as if installation is a daily occurence (I damn sure hope not) -
But in any case, sysinstall is getting dated.
Yeah. Plus, it doesn't support ZFS, so it's going to have to be changed anyway.

Oh, and to those who say "use PC-BSD/DesktopBSD": first, DesktopBSD is still based on 6.3, and second, there's a pretty good chance that KDE isn't what user x is looking for. There's a big gap between a default autosetup OS with KDE and one with an ncurses-based installer. finstall doesn't have to come at the cost of other features; it's not a super huge project.
Also, as a final thought: perhaps the GUI frontend for finstall could be based on a lighter-weight toolkit than gtk, like fltk--would that cure concerns about "bloat"?

*in other words, many people don't have the time for
 
I didn't expect to see this thread pop up again. I thought the issue was laid to rest.

I think people are completely ignoring the issue this thread addresses.

It addresses the support and sometimes total dismissal of the finstall project to move into production.

The last couple posts are talking about DesktopBSD.
1) This is not FreeBSD and not really relevant.
2) DesktopBSD uses finstall or a variant thereof (I have not
installed DesktopBSD to verify exactness)

So basically we are confronted with the idea that other implementations of FreeBSD into other distributions *all* seem to forgo sysinstall and instead select finstall in its stead.

The point of whether it uses 6.3 or 7.1 or 4.11 as it's base is essentially irrelevant (so although the one poster was slightly behind on the status of the DesktopBSD project, his point wasn't over the version snapshots).

I hate to beat a dead horse over and over again (<sarcasm>as much fun as it is... </sarcasm>) but I feel that when people make attempts to "chime in" to posts they completely neglect the thread's origins and the merit of the discussion at hand. Fact-filling irrelevance does nothing but clutter the thread with many different debates. I think we are back on cue as far as the gui installer idea?

I guess as a side note that we appreciate attempts to "fact-fix" information. I personally don't care about extraneous info unless it affects the discussion at hand.

I almost just deleted this post before submission and leave the thread as is. But I did type this up and so we can all take it or leave it. Enjoy.
 
robertclemens said:
I didn't expect to see this thread pop up again. I thought the issue was laid to rest.

I think people are completely ignoring the issue this thread addresses.

FreeBSD offers a bootonly CD for installation. A good thing cause it's only 30MB small and all other stuff comes up to date via FTP so there is no real need for downloading megasized installation DVDs.

Bad thing is, it uses sysinstall and that has problems with resolving the ftp-servers hostname making the the CD worthless. An example can be found here http://forums.freebsd.org/showthread.php?t=2087. But up to now that issue does not really interest anyone...
 
Erratus said:
Bad thing is, it uses sysinstall and that has problems with resolving the ftp-servers hostname making the the CD worthless.

Yet another complaint about sysinstall...

I can't promise anything yet, but it just might be that during the summer I'll have the time to rework that thing. Should be a fun project to do.

Alphons
 
robertclemens said:
I didn't expect to see this thread pop up again. I thought the issue was laid to rest.

I think people are completely ignoring the issue this thread addresses.

It addresses the support and sometimes total dismissal of the finstall project to move into production.
Honestly, if you'd stop trolling the thread it might be laid to rest. The question was answered a few pages back as to why it we don't have one.

Basically it comes down to:
A) We have a GUI named Sysinstall that works pretty well.
B) The people doing the work have other priorities.
C) People that can't handle Sysinstall, even with all the help, are unlikely to be able to figure out the more complicated challenges after installation.
D) There's a tremendous amount of work that has to be done for an install program like finstall to work properly and consistently, that takes time and energy. Getting it wrong could very well leave us worse off than where we are at the present.

Nobody is seriously suggesting that it wouldn't be nice to have a more pretty and up to date install program. But it's difficult to do that and have something that'll run properly on the older hardware that's still supported.
 
hedwards said:
Honestly, if you'd stop trolling the thread it might be laid to rest. The question was answered a few pages back as to why it we don't have one.

Basically it comes down to:
A) We have a GUI named Sysinstall that works pretty well.

Hmm, perhaps you did not read all messages. As you could have read sysinstall does not work pretty well!

It is by no means constructive to call other opinions a troll.
 
Erratus said:
Hmm, perhaps you did not read all messages. As you could have read sysinstall does not work pretty well!

It is by no means constructive to call other opinions a troll.

Can you be more specific? What is wrong with sysinstall in your
humbled point of view?

I am not big fun of sysinstall myself. I think it is way too complicated and non flexible. I personally would like to see something simpler like OpenBSD script so that I can actually boot FreeBSD machine with the single floppy disk or in the worst case scenario something like NetBSD installer.

What would you like to see?
 
Immortal thread, eh?
I said it before and I say it again. If you are new to FBSD and sysinstall is not working for you do manual install as your first install. Your time will not be wasted, on the contrary, you will know your way around after that and it will save you time and headaches for rest of your life.
Next time you try sysinstall it will work for you, guaranteed. ;)
 
This is not the point. Not using it is not a solution to its problems. Barely any experienced user uses sysinstall.

This should tell us all something about the quality of its interface.
 
Hey I don't find it bad. Actually I in fact like it.

However it does need to be updated and extended, namely regarding letting one use gjournal, zfs, etc.
 
kamikaze said:
This is not the point. Not using it is not a solution to its problems. Barely any experienced user uses sysinstall.
Obviously I do not qualify as an experienced user then. I use sysinstall to install the base system, then I tweak my make.conf, rebuild the world and proceed with installing software from ports. At this point I could remove it from system because it is not needed any more. For this purpose sysinstall works great for me.
Sysinstall = install the system ... right?
Am I missing something here?
 
I find the sysinstall is an easy way to have a working platform for install. If you have suggestions how to do the same without sysinstall - I'm all ears!
 
Wait a minute ... (I do not pretend to be an experienced user, as you rightfully said)
I obviously want to sync the sources first. So it would be like this:
prepare HDD
sync (get) sources
set up make.conf
make buildworld
proceed with what you said
 
Nope, I had a chance to think about it. It's not that easy, I'm going to miss lots of stuff in /etc which I have to create manually, etc.

May I ask how you, an experienced user, install FreeBSD?
 
As was pointed out earlier in this thread, this whole discussion isn't basically about GUI or Text mode, but rather about usability. And when it comes to usability, sysinstall is obviously neither the best choice, nor the worst.

Maybe one needs to rethink the basic goal of an OS installation routine. As I see it, the basic goal is to guide you step by step through the installation process, telling you exactly what is going on, and requesting the information that it needs to fullfill it's task - installing the operating system, while neither overwhelming unexperienced users with lot's of options they don't have the least inkling about, nor keeping experts from choosing the funky options they like.

Having that said, I see two different kinds of problems:

  1. sysinstall in it's current form is not really step by step, but rather entering a big menu of things, giving you this 'now what?' feeling. The navigation, using TAB/SPACE/ENTER is really bad, especially when it comes to the OK/CANCEL buttons, or returning to the previous menu. There are places, where sysinstall provides little to no information to you about what it is expecting you to do, or what kind of information it needs from you in order to continue.
  2. The user has to provide some information. Not being able/willing to provide this information ultimately leads to not being able to install the OS.

The first problem is a sole software problem that can be fixed. The second problem however cannot be fixed, cause there is a subtle difference between a newbie and a person strictly unwilling to learn anything while expecting everything.

To pick up the comparison with a car someone brought up earlier: If you just wanna drive your car, not knowing wether it will take diesel or high-octane fuel and where to refill it, you probably wont get far.
 
Back
Top