“Oldtimer Linux gurus get rather cranky about graphical interfaces. Sissy, bloated, slow. If it is too easy it is suspect, unworthy of the true sysadmin. I think it’s nice to have a rather more flexible perspective, and to rejoice in the wealth of choices.” Read the article at EarthWeb.
While hardly a guru, I have been using linux exclusively for a few years now. Some of the GUI stuff is nice for quick and dirty changes, but no GUI will ever offer the total flexibility if the command line. I like using GUI’s for some stuff. But when you take the time to learn and work at CLI, DAMN. It amazes me everyday how quickly you can get things done with a few stringed commands of a little script.
Just my take, but I never again want to be limited to GUI.
Bill
I like my UNIXes headless. In my experience, X is a bloated and un-fun shell, kinda like Windows 3.1 on steroids. That’s not to say it doesn’t have a use; I kinda like using Mozilla 1.1 under X-Windows, for instance. But when it comes to running a server, I find X-Windows to be a complete waste of resources. My website’s server is far, far too overloaded as is
(Pentium 166MMX with 32MB RAM. Do you _really_ want to justify running X-Windows on that, along with Apache, Postfix, MySQL, Posadis, et al? Headless FreeBSD cuts the mustard quite nicely. Although it could still use a bit more RAM – anyone have some SIMMs to spare?
Well its standard practise to install the minimum on production servers anyway. It just makes sence .
I know some people who would do the equivalent of a RedHat ‘everything’ install on their webserver, quite a lot of bloat indeed, something with running X and having compilers on a production server scares me, personally I just install the absolute bare minimum to keep the bloat down.
As for installing X, some people are forced to do it because they only know how to use some of the X config tools :?.
As for the P166, X/kde runs slow as crap on my 1.2Ghz duron, let alone X on a 166.
How does a minimum install of FreeBSD compare to Linux on that box? Interested to see which performs better on older platforms.
When starting out with Linux, trying stuff I still had to learn, I was amazed by the possibilities provided by Webmin, which you could call sort of a GUI.
But when I was learning more and more, I realized I’d rather manage things by CLI, wich indeed is more flexible.
Besides, CLI is ‘universal’. GUI stuff is so specific (Distro-specific, and indeed limited) that you just can’t do everything everywhere.
Besides, when managing servers, which we usually always do remotely, it is pretty clear that an SSH shell will always do the job ๐
Maybe slightly off-topic: some time ago I attended a seminar offered by NetIQ and MS. One of the MS guys, a technical guy who told some marketing stuff about upcoming dotnet server, said this new Windows server would be totally manageable by command line! I wasn’t able to find any resource confirming this ๐ Maybe someone can?
>>a technical guy who told some marketing stuff about upcoming >>dotnet server, said this new Windows server would be totally >>manageable by command line!
Are you sure that was not a marketing man?
I prefer Slackware on my server @home but in a business envoirment RedHat is still them most standard gnu/linux distro. I still hope they will put more effort in stabily and preformance rather then gimicks and desktop.
Bas
in RH 7x, the minimum is around 300 MB
it took me a long time to strip RH 7.2 down to 70MB
for my AMD 586-133 with a 170MB hard drive ๐
by the way, RH 8 choose UTF8 as my default language,
truely a software for the global audiance – a bit
of a mess for non utf8 terminal for viewing man page
glad that they include sshd in the default install
And I expect them to get better.
RedHat Linux will give both Apple Computers and Microsoft a run for their money!
The command line will always be there as well. If that’s your cup of tea, go for it! These GUI tools will go a long in getting Linux first class status in the corporate world.
ciao
yc
>One of the MS guys, a technical guy who told some
>marketing stuff about upcoming dotnet server, said this
>new Windows server would be totally manageable by command
>line! I wasn’t able to find any resource confirming
>this ๐ Maybe someone can?
Awhile ago, I ordered the Win.NET RC1 preview thingie from MS. One of the things that was mentioned was that a bunch of UNIX guys at MS had worked a lot on .NET server, and one of the things they added was the ability to fully administer the server from the command line. Supposedly, every gui tool now has a CLI counterpart.
Also, the new IIS 6 has an xml metabase (i.e. text config file) that can be updated on the fly. As soon as you save changes, IIS changes its behavior. Seems like MS is taking a few cues from *nix.
Also, the new IIS 6 has an xml metabase (i.e. text config file) that can be updated on the fly. As soon as you save changes, IIS changes its behavior. Seems like MS is taking a few cues from *nix.
Except with IIS6 you don’t have to do a apachectl -restart or whatever the command is
Personally, I think GUI configuration tools are a good thing to have. Obviously, if you don’t want to use them, you don’t have to. However, they make a nice set of ‘training wheels’ for people who are just starting out, allowing them to get things done while they are learning their way around the CLI (ie – taking off the ‘training wheels’). IMHO, it’s always nice to have the option to use the GUI tools if you want.
Only bad thing about GUI tools (as someone has already pointed out) is that many of them tend to be distro-specific, so what works in one distro may not work in another. However, I guess you could say the same for some CLI tools as well. How many of the nifty CLI utils that come with Slackware (ie: the ones that allow you to configure your network, switch default window managers, etc) are available in other distros?
I don’t think Redhat’s GUI networking tools are good enough or even very functional. Xp networking is still a lot easier.
Of course, no serious unix admin wants to run an X Server on a production server. Why would you want to do that? Its not only a waste of resources but of your admin time as well. Writing a small script to setup your virtual hosts, for example, is 100 times easier and faster than using the Redhat GUI Apache tool. I personally see GUI tools as a way to entice the MCSE crowd.
>>Of course, no serious unix admin wants to run an X Server on a production server
Say what? Why not? A good unix admin will size a server properly and use what ever services are needed.
Not all Unix servers are simply used web, ftp, db and mail.
There tons of very rich business and engineering applications that use X.
If one doesn’t need X then… but then again that doesn’t necessarily make one a serious unix admin either…
ciao
yc
> How does a minimum install of FreeBSD compare to Linux on that box? Interested to see which performs better on older platforms.
I don’t know about FreeBSD… it’s said to be even leaner…
Anyway, I use Linux on a Pentium One 133, 32MB RAM.
It works well. Of course, no KDE and no Gnome as they’re too heavy for such old machine. XFCE 3 runs ok.
I also use a lightweight window manager (Icewm, Blackbox, flwm, I switch all the time).
Mozilla is impossible, too, so I favour Opera and ol’ Netscape 4.x. There’s a lot of other excellent browsers… I specially recommend Links (look for it at freshmeat).
For word processing, Openoffice won’t run, I used Wordperfect 8 instead. But, over the last year I switched to Abiword and now, with tables, I think wp8 will be unistalled.
Gnumeric works, albeit slowly…
For mail needs, Sylpheed rules. It’s very light and there’s a “claws” version with edgy features. Works like a charm.
I play Quake I, or more exactly Team Fortress. Quake II is too slow here (I don’t have a 3d card). I got a very low-latency line, the machine is not a problem (but image quality sucks, I admit).
Music is no problem, but (real)video is very limited.
A curiosity about open source/free software: as time passes, better software runs on the same hardware.
>And I expect them to get better.
>RedHat Linux will give both Apple Computers and Microsoft a run for >their money!
I wasted my weekend configuring RH 8 to share connections with my other computers. I had all sorts of DHCP problems, even my modem was not found and had to do some more tweaking.
Things like that are configured very easy in both XP (a wizard) and OSX (single click).This is also much better done in Mandrake.
Linux still has a long way to go to be easy to use. I’m not even sure why people want Linux to be easy, people who use Linux don’t need ease of use, otherwise they would chose an alternative.
>>people who use Linux don’t need ease of use, otherwise they would chose an alternative.
HeeHeee, good one!
You must admit however that it saves money!
Imagine rolling out a small web farm, with database backend, email etc…
NT Software for each box may run you several thousand dollars.
Linux Software cost is an absolute $0.
You gotta love it.
ciao
yc
the easiest networking setup procedure i’ve ever encountered was GNU/linux mandrake 8 on a pentium pro 200 box, compared to getting the thing running with winNT4(hell on earth, mylex RAID anyone?) or nt2k neither of which would work correctly without extra 3rd party drivers, all of the hardware was detected and set up correctly including the RAID board and it’s 3 drives, network card(3com of some type) worked nicely, network setup consisted of give it an IP, host name and DNS server and it works, could then use samba to log on to the NT domain and use network shares, very nice, very clean, very quick.
it will test your disks before starting the installation. I suspect it’s merely running a checksum – what does she think a checksum is?
A peeve dear to my heart is the lack of an option to set the hostname.
I have not heard of any *nix ever not configuring the hostname during the install, and I have installed a *lot* of *nices
KickStart – on my RedHat8 install, although fortunately tftpd is not running, the images are there in /tftpboot – this is not something I want or need.
Firewall – If you’re prepared to configure your own firewall, do so. If you’re not, do you really care whether it’s ipchains or iptables?
Sendmail – Again, Carla shows that she does not know what she is talking about. Sendmail provides two services – that of receiving and distributing mail, and also, surprisingly, considering its name, sending mail. Hmm, why would sendmail be installed? Maybe because it allows you to send mail?
text editors and terminals are buried deep in the system menus. I don’t care how desktoppy a user is, those are basic Linux tools, and should be front and center
Make your mind up! The article starts by belittling the CLI as only-for-bearded-gurus, and ends by complaining that text editors and CLIs are a whole Start/System Tools/ away?
Remind me to hire Carla for my next Unix install….
yc: Not all Unix servers are simply used web, ftp, db and mail.There tons of very rich business and engineering applications that use X.
But that’s not a server, that’s a desktop.
I’ll be generous and assume that you mean a firewalled, internal server, which has X installed, which users can ssh to, and run their graphical application from, whilst xhost’ing them to Exceed, or another *nix desktop. Even then, you do not need X to be running on the server. It can boot into runlevel 3, have no graphics card or monitor installed.
So what was your point?
Steve
“I have not heard of any *nix ever not configuring the hostname during the install” Dear friend, in Red Hat selecting DHCP during installation means no way to set the hostname during install. If you assign a static IP then you have the option of setting the hostname. Score: -3
“Firewall – If you’re prepared to configure your own firewall, do so. If you’re not, do you really care whether it’s ipchains or iptables?” Why, I suppose you could even mix n match! <–sarcasm. For many users, the simple ruleset created during install is sufficient, especially if they are not running any publicly accessible services. It certainly matters when you want to make changes, if the rulesets are in ipchains or iptables. Score: -5
“Sendmail provides two services – that of receiving and distributing mail, and also, surprisingly, considering its name, sending mail. Hmm, why would sendmail be installed? Maybe because it allows you to send mail?” No, ignorant man, Sendmail is an MTA. It is not necessary for sending and receiving mail when have an email account on an ISP, or use a service like Hotmail or Yahoo. Sendmail is needed only if you are running a mailserver. The article said Sendmail is installed even when you install Postfix, which it should not do. Score: -10 for ignorance of basic facts.
“text editors and terminals are buried deep in the system menus. I don’t care how desktoppy a user is, those are basic Linux tools, and should be front and center
Make your mind up! The article starts by belittling the CLI as only-for-bearded-gurus, and ends by complaining that text editors and CLIs are a whole Start/System Tools/ away?”
My dear Steve, you are not even a good troll, you are a very poor troll. The article does not belittle CLI at all, it presents both means in a positive light. I agree with the article, that choices are good. Score: -25
Total Steve score: -43
You just don’t want to like this article, or anything.
I’ve seen comments rather like this one many times in the past:
While hardly a guru, I have been using linux exclusively for a few years now. Some of the GUI stuff is nice for quick and dirty changes, but no GUI will ever offer the total flexibility if the command line. I like using GUI’s for some stuff. But when you take the time to learn and work at CLI, DAMN. It amazes me everyday how quickly you can get things done with a few stringed commands of a little script.
Just my take, but I never again want to be limited to GUI.
Limitation is not an instrinsic property of the GUI. A poorly designed GUI system is limiting, but so is a poorly designed CLI system. Power isn’t inherent to a CLI. I’ve not been able to put my finger on why people think so, but my best guess so far is that because there is usually a higher learning curve for the average CLI, users like to feel that this added time in learning how to use the system is justified by more “power.” In many cases, the more work and more time spent learning is justified. However, oftentimes, this isn’t because the CLI is entirely more powerful than the GUI, but that the CLI provided on whichever particular system this user is using has different strengths than the GUI on that same machine.
Most users extolling the virtues of the power of the CLI over the GUI are veteran Windows users discovering Linux -or- current Linux or Un*x users who find the Unix shell system power powerful than what is available on top of the X Window System currently. In those cases, I can’t argue with the point that bash
For the sake of discovery, let’s look at what most folks are thinking/talking about when they compare the power of the CLI with the GUI. For the GUI, most users are thinking of what they are used to using on Windows or X11 (including GTK+, Qt, Xt and Motif). The CLI is usually a reference to bash, or one of the other POSIX shells. Windows is often inconsistent, and X11 is notoriously inconsistent, even when using a single desktop, like GNOME or KDE. In these GUIs, you if there’s something you want to do with the data that isn’t a part of the GUI, you’re probably out of luck. Some applications have been endowed plug-in architectures, but not surprisingly, there is no consistency in that- there’s no standard whatsoever. Here, there is a glimmer of sanity- AppleScript on Mac OS Classic and X, and CORBA for GNOME. Both the AppleScript and CORBA/GNOME systems allow a wide variety of computer languages access. Outside of that, there’s very little consistency.
On the otherhand, POSIX shells are mostly consistent. Not as consistent as they could/should be, but still better than your average suite of GUI apps on a Linux system. You can easily redirect data between CLI programs and files, for cold storage. This data, which is being passed around and being manipulated represents something that you often can’t get a hold of in a GUI app- some raw form of the data your application is dealing with. So, in a sense you’re getting at the innards of the application. This data can be passed using pipes to programs written in any language. However, POSIX shells leave a lot to be desired in many areas and very limiting, though often in other areas as a poorly designed GUI system.
The DOS (command.com) shell is a great example of how powerless a CLI can be.
I hate to always bring it back to Squeak, but it is a great example of a GUI system that is very powerful, reflective, extensible and easily manipulated without having to write and code. To prove that it’s not the only one, there is also Oberon (an OS based on Oberon was mentioned here semi-recently) and Self. Morphic, the GUI framework of Squeak and Self, is easily more powerful than any Unix shell. There are other systems as well. The Mac OS has a very powerful GUI system, thanks to consistency and the well designed AppleEvents/AppleScript system. However, it could go farther and be more pervasive.
Just another rant that no one is probably reading…
I have not done a DHCP install of RedHat 8 – I would assume that it gets a hostname from the DHCP server along with the IP address (that is, after all, what the protocol specifies). If it does not, then that is a clear, simple bug that should be filed with RedHat.
As you say, either iptables or ipchains would be sufficient for a typical user, and the rules installed by RedHat are adequate. So I question why the author has such an issue.
Many applications are hard-coded to send mail via sendmail, so having the binary installed, but the daemon not running, makes a certain amount of sense. Sendmail is not “needed only if you are running a mailserver.”
Whether or not the article is belittling may be down to personal interpretation, but saying “Oldtimer Linux gurus get rather cranky about graphical interfaces. Sissy, bloated, slow. If it is too easy it is suspect, unworthy of the true sysadmin.” doesn’t sound very positive to me.
> Limitation is not an instrinsic property of the GUI.
Actually, it is. A GUI requires a human (or at least a Microsoft living organism). If you want to configure 100 servers using the GUI only, then you have to click the same settings 100 times.
However, with a CLI / API interface you can automate this process and do the work only once. Furthermore, simply saving the shell script saves the settings since it can be used over and over.
I was once at a client where they were opening a new office with about 500 users. The manager provided IT with an excel spreadsheet of the new users and about 5 NT admins set about “clicking the gui” to set them up. They would have been there for days. I took the spreadsheet and in about 1 hour had a perl script that got them all in.
Actually, it is. A GUI requires a human (or at least a Microsoft living organism). If you want to configure 100 servers using the GUI only, then you have to click the same settings 100 times.
Actually, no. A GUI requires a human no more than a CLI does. If you want to copy a 100 different files using only the CLI only, then you have to type a command with each file name 100 times.
Oh, you say a script could be written? Scripts can be written for GUI systems as well. As I said many times, poorly designed GUIs impose limitations, just like poorly designed CLIs. There are scriptable GUIs that can achieve the same amount of automation as a POSIX shell.
I was once at a client where they were opening a new office with about 500 users. The manager provided IT with an excel spreadsheet of the new users and about 5 NT admins set about “clicking the gui” to set them up. They would have been there for days. I took the spreadsheet and in about 1 hour had a perl script that got them all in.
I have done similar things with GUI systems. Automation is nothing new, and it isn’t limited to a command-line interface.