Page 3 - Linux RedHat VS FreeBSD

Ok, freebsd, I'll bite.

As a relative newbie to compiling and setting up server software and the like, I'll admit, I don't have much experience in this. The lack of direction in Linux as far as where to put stuff in /usr/local/ has kind of bothered me.

Let's say I'm interested in downloading and compiling stuff like Apache and MySQL rather than using rpms. What are you referring to as far as PREFIX that should be used, beyond /usr/local, to prevent having /usr/local/apache, /usr/local/mysql, etc.?

Thanks,

nuk

>> What are you referring to as far as PREFIX that should be used, beyond /usr/local, to prevent having /usr/local/apache, /usr/local/mysql, etc.?

No beyond of /usr/local .. /usr/local is the best and right place to prefix to get your apps more stable and less problems as well. If you don't like it then stick to Linux..

I'm not saying that I at all disagree w/ what freebsd (or you) are saying about the organization of /usr/local; just the contrary. I've always been put off by the lack of direction in /usr/local, and want to know what you think is the 'right' way as far as how things should be organized in /usr/local, regardless of whether its on FreeBSD or Red Hat Linux.

Put more simply, I guess I'm asking, how do you do it so that you don't end up w/ /usr/local/apache, /usr/local/mysql, etc.? Unpack everything under /usr/local/src, and then use a --prefix=/usr/local option for ./configure to have it put the binaries in /usr/local/bin, etc.?

Thanks,

nuk

Typically when your build applications in FreeBSD, they are referred to as ports. So, you really dont have to go around creating directories, etc. When you build a port, it creates all that for you. It will default to /usr/local .... whatever.

Just curious, why is this an issue for you ?

EDIT: you just do either "make followed by make install followed by make clean" or just make install clean, assuming default configurations, it will retrieve the sources, unpack, configure, compile, install and clean! It's that simple.

Yeah, you are correct about unpack in /usr/local/src, compile by prefix to /usr/local .. When you are done, then you can remove the Apache source code that you unpacked.. Just read freebsd's post about it at http://forums.devshed.com/showthrea...4897&forumid=15 ... You also can search for prefix under freebsd's nick/name.. He explained about it so many times, so search for under his name is worth to learn more stuffs from him.

In FreeBSD, I don't really have to worry about it since it's already set by default.. I usually modify in the Makefile and freebsd.layout to change the docroot point to /www instead /usr/local/www.. FreeBSD does the most job for us..

Because I'm running Red Hat Linux (or any Linux, for that matter, same issue) and am in the process of rebuilding my server for my home LAN w/ the latest kernel, LVM, and then installing Apache, MySQL, etc. for educational purposes (this server normally just provides services like dns, dhcp, nfs, samba, squid, etc.).

I have three main paths to choose from:

a) use the vendor-supplied pre-compiled binary RPMs which will put the software under /usr, but I'm stuck w/ whatever whoever packaged the software deemed appropriate for compile time options. Will I notice a big difference in performance in my situation? Most likely not, but I'm not sure I'm 110% comfortable w/ this.

b) obtain the latest stable versions of the software I want, and compile it myself under /usr/local, customize the compile-time options to my hardware and situation, and hopefully learn something in the process. Linux doesn't give a whole lot of direction as far as this goes, hence my asking here.

c) take the vendor-supplied source rpms and recompile them to my architecture. Hopefully this would get me some more performance (like I said, not a big deal) but still leave things in a 'standard' (for Linux) location. Have to be careful of future upgrades, though. Optionally, I guess I could theoretically either 1) take the source rpms and tinker w/ the spec files and get the compile options I want, including location, to protect against future problems w/ rpm or 2) build an rpm from the source tarball, and set it up the way I want. Unfortunately, both 1) and 2) are a bit beyond my current capabilities.

Thanks,

nuk

Thanks freebsd for your answer!

Seems like Net is something for me to look into then. I've been using Free and Open for a couple of years, so I think I have the fundementals down. Or so I hope at least.

Thanks again.

/Fjodor

best is to compile with PREFIX=/usr/local

Maybe not if you do know other dists of linux, which i don't. I tried with slackware and debian once, long time ago, maybe I should try them again?

Linuxconf is awful. It had corrupted my dns zones and also mail configuration by altering the standard sendmail cf's (or mc's? This is insane). After wondering for weeks in search of a flaw that made sendmail behave weird (Slow response time for example), I just tried installing Redhat 7.2 and configuring it myself by editing configuration files manually. Voila! All email probs vanished.

No wonder linuxconf is not installed on 7.2 by default.

Anyway, installing X on a production server is not wise. In fact, its not wise to install anything besides the service itself, and barebone necessary bins/sbins. This, of course, as a typical security precaution. Some fellas doesnt even install a compiler on servers.

If you want a graphical type admin tool w/o installing X and all it's related dependencies, try Webmin. It uses perl scripts in the form of 'modules', so you can install/uninstall modules as applicable to your system. i.e., if you have PostgreSQL and not MySQL, then simply uninstall the MySQL admin module and forget about it.

It uses, IIRC, perl to accomplish its tasks, and from what I've heard/seen, does them in a way that doesn't mess w/ you editing the files by hand like linuxconf did i.e. you can make changes by hand, and webmin will honor the changes and not blow them away.

It uses port 10000 by default, and can be configured to use OpenSSL, so the connection would be as tight as any other secure web connection.

Best of all, it's not platform/OS dependent, so you can use it on any flavor of Linux, BSD, etc. and keep a consistent interface on things.

Is it easier/quicker for an experienced person to just ssh in and fire up vim and edit a config file? For all but the most trivial changes, probably. Does webmin save any administrator from common typographical errors for routine operatiosn? Yes. It gives novice administrators a somewhat more friendly interface than a terminal screen, though I will concede that sometimes the depth of options included in some webmin modules is intimidating if all you want to change is one simple thing (some modules list every option, even the ones w/ default values, which can be a lot of them!!)

FWIW,

nuk

Webmin? No way for a security-wise person.

>> can be configured to use OpenSSL

Implementing SSL means the connection is encrypted and it's just a part of security concerns. Webmin is still insecure in other aspects, just to name a few: tons of local exploits and DoS. So you can't really assume, yeah I run Apache + mod_ssl, therefore it must be secure.

>> webmin save any administrator from common typographical errors

Maybe. However, like linuxconf, webmin breaks things more than time-saving. The qmail module for instance, it's totally broken and is unable to parse /var/qmail/users/assign file correctly.

>> It gives novice administrators a somewhat more friendly interface

A newbie should learn how to administer the system correctly and preferably without X and GUI before relying on something (webmin in this case) that can't really be trusted because of a poor security record. If this newbie can't survive without X, just go back to Windows.

Like in FreeBSD, I don't trust sysinstall not because it might not be secure but it breaks things from time to time.

I guess I am late on the discussion, but I have been working with FreeBSD for over 6 mos now, and find it powerful and straight forward. Infact, I got fed up with my windows machine, so I am running FreeBSD as my main desktop now w/KDE.

Anyway, I am of the opinion that any linux distribution out there is a lot more convoluted and hacky in the way it is set up. The standardization in FreeBSD allowed me to quickly pick up how the system was structured.

With that being said, I would say that FreeBSD is definately superior to RH (and any linux distro) in its usability. I disagree that FreeBSD is "just for experts", etc. I am no expert, nor do I claim to be, nor did I graduate from linux. I would recomend FreeBSD as the first and only alternative OS to anyone willing to try one -- newbie or self proclaimed *nix god.

As I said, I am no expert in the internals (kernel, code, etc) of either linux or FreeBSD, but it is my general impression that FreeBSD is again far superior. I attribute this to its roots, maturity, development process, developers, and its user community.

Brett

If someone asked me that question and I thought they were serious. I would not even recommend Linux before BSD. If someone asked me that and I didnt get that they were serious, I would tell them to try Mandrake. But when you ask the question "Linux Redhat v/s BSD for a webserver"....I think the question itself is kinda broad. I mean what kinda site?....ssl?...php?...?database?... how much traffic? reply with this and I'll give you a serious answer.


Its is excellent, if you dont want pf use ipf:

http://openbsd30.ipfilter.org/

Its stability and reliability have never been a problem in the past. And if you think that it is now b/c of pf then use ipf. Problem solved I hope. So please, at ease the trash talk about OpenBSD. It really is an excellent choice for an operating system.

mmm, and all this time i thought that linux was one of the best Os's on the market!
Everyone seems to be talking about RH and how they mistreat the file system, but what about Mandrake linux, do they to screw it up?
And what about FreeBSD? I am correct in saying that it is an exellent server and not so good a client machine? cause at work, upon my recommendation, we have apache,php, various communication gateways and mysql running on a quad processor system using Mandrake, but i gather Mandrake is frowned upon, how would BSD make it better? that is apart from any security settings it might have setup because we basically reconfigured security settings on our server.
Is it really that hard for a linux user to migrate to BSD once u know know unix basics and are very good with linux?
cdburgess75, in response to ur posting, the server that we are running at work is very heavily loaded "e-commerce" server, with apache, ssl, php and mysql all running on the one quad processor server (we need to save money so we built our own server and made it as fast as possible!), which would be better then? Mandrake thats already running? or a BSD? and why is BSD so much better? all this time i thought that linux was a great system, is it really this bad??


Oh yeah, 1 more thing, after reading about this post i am very curious about BSD, apart from security, how does it differ from the way linux handles things? cause linux has never crashed on my in the 2 years i have being using it.
no huge explanations needed, just a weblink or something small. thank u

1 final thing, i am up to date with linux kernel developments, any features in freeBSD's kernels not it linux and vice versa? (including the hyped 2.6 kernel for linux)