Which is your Favourite Linux?

[jdhore]

To put the bloated piece of sh!tware known as Red Hat into perspective by looking at the US-CERT NVD database:

          3-months      3-years     All
Windows   113           1429        3070
Red Hat   248           2228        5869
FreeBSD   1             44          443

Red Hat is probably the WORST operating system one could be using and couple it with the inferior architecture of i386 and you have a Frankensteinish POS aching for disaster.

I'm sorry, but this argument is total crap.

For FreeBSD and Windows, it only counts the "core" components of the OS (the kernel, OpenSSL, FBSD [*] FreeBSD libc, tcsh, etc).

For Red Hat on the other hand, it counts all ~3,000 packages in the repositories (and that's just for RHEL, not even counting Fedora which has ~22,000 packages in the repositories. Although maybe the wikipedia numbers are incorrect).

That's a lot more software than the FreeBSD base system to have security issues (that includes a bunch of HTTPd's, a bunch of SQL engines, a bunch of programming languages aside from C (Ruby, Python, Perl, etc)

 

[Simba7]

inferior architecture of i386

..and what exactly is wrong with the i386 architecture? I've been using it for decades without issues. I still operate a set of i386 (well, i686) boxes along with a set of 5 SunFire V100s and they all are rock solid with FreeBSD and Gentoo Linux.

 

[SR_Ind]

..and what exactly is wrong with the i386 architecture? I've been using it for decades without issues. I still operate a set of i386 (well, i686) boxes along with a set of 5 SunFire V100s and they all are rock solid with FreeBSD and Gentoo Linux.

This argument flows out of the old RISC vs CISC debate. In university we were always told RISC is superior (hence i386 is crap). Reality has turned out to be otherwise. In the desktop world the fate was decided long ago (sealed the day Apple joined the camp).

On the server side RISC is fast losing out, in fact old UNIXes running on RISC processors are getting kicked out of the market. For example look at HP-UX's market share for last few years, also try to find sales figures for Solaris on SPARC vs x86.

I don't know how many people here do commercial software development, but last month one of our customers, a large (maybe the largest?) European telecom equipment maker told us that on the managed networks they manage for operators worldwide are only allowing x86 hardware. Critical OSS/BSS application on older servers are gradually migrated. Only non-essential applications shall remain on non-x86 platforms.

Weird, but they have better insights into the market and technological trends.

[*] [*]

 

[gore]

This argument flows out of old RISC vs CISC debate. In university we were always told RISC is superior (hence i386 is crap).

Reality has turned out to be otherwise. In desktop world the fate was decided long ago (sealed the day Apple joined the camp).

On server side RISC is fast loosing out...in fact old UNIXes running on RISC processors are getting kicked out of market. For example look at HP-UX's market share for last few years, also try to find to sales figures for Solaris on SPARC vs x86.

I don't know how many people here do commercial S/W development, but last month one of our customer, a large (maybe largest?) European telecom equipment maker told us flatly that on the managed networks that they manage for operators worldwide, they are only allowing x86 hardware. Critical OSS/BSS application on older servers are gradually migrated. Only non-essential applications shall remain on non-x86 platforms.

Weird, but they have better insights into the market and tech trends.

OK, now I HAVE to pop in here for this one; It wasn't easy to understand exactly what you meant, but that's OK, I think I got it.

I personally don't know how ALL of YOU feel about this, but even with the way things are going right now, where regular PC Hardware is becoming the norm, if I were in charge of some huge company, and I was told to set up our "Mission Critical" section, I probably wouldn't be choosing regular off the shelf stuff...

Call me crazy, but an SGI Server, and SGI Workstations, are better than PCs. Even though SGI is going to a Linux only World, which sucks because IRIX is Beautiful, I'd still want that. Sun, The Sun has Set, and they might as well be renamed "Moon" and change SunOS to MoonOS (Which in a twist of humor is a version of Linux! LOL) but the reason is Oracle is ruining Sun.

If you've ever seen "25 years of Berkeley Unix" where Marshal Kirk McKusick is talking about how he saw /dev/kidney and they were using Unix for Kidney Dialysis, and he says "I sure Hope I don't ever need that" I agreed with him. If I had to pick something where crashing could kill me, I'd want either the SGI stuff I just brought up, or FreeBSD.

I'd sleep better at night knowing I had some nice HP Servers that didn't know what Windows was lol.

 

[throAU]

RISC vs CISC is a bit blurred these days.


CPUs are complex enough now that the intel stuff is all RISC internally with a decoder for CISC instructions.

So you get the benefits of CISC style code density but the code is broken up and run as RISC internally. Also, the RISC CPUs got more complex and added streaming instructions like altivec...


I ask why is an SGI or SPARC cpu so inherently superior to x64 these days? Sure in the past they had the memory bandwidth, etc - but intel has the best fabs in the world, and individual x64 nodes are now fast enough to blow basically anything else out of the water in terms of instructions per dollar with cheap clustering solutions. The intel gear has virtualization support, redundant hot swap hardware, etc. The old advantages simply either are not there, or are way more dollars than they are worth, and can be worked around by just throwing money at x64.


But it will come full circle. Eventually just like intel ate the server market from the bottom up, ARM is going to eat intel's market from the bottom up as well.

 

[kpa]

I386 and amd64 survive only because of compatibility requirements and the fact that it's infinitely much easier to find working development tools for them than some obscure RISC architecture. From a completely technical point of view i386/amd64 stink to high heaven.

 

[throAU]

I386 and amd64 survive only because of compatibility requirements and the fact that it's infinitely much easier to find working development tools for them than some obscure RISC architecture. From a completely technical point of view i386/amd64 stink to high heaven.

Be that as it may, they have the horsepower. At the end of the day, 99.99% of developers are not dealing in x86 assembly language any more. So the "crap" architecture doesn't actually matter.

 

[SR_Ind]

Be that as it may, they have the horsepower. At the end of the day, 99.99% of developers are not dealing in x86 assembly language any more. So the "crap" architecture doesn't actually matter.

x86 rules the world of personal computers (almost complete domination) which translates to it being the software development platform as well. That means it is easier (and for hobbyists only possible platform) to target x86 for software products.

The second factor is the availability of the non-x86 platform. Non-x86 platforms are neither cheap nor commonly available as personal devices. So how does one write software for them? Hobbyists and FOSS developers are ruled out, small (in developing countries even mid size) companies are ruled out. All of these folks are a big mass of software developers that have contributed enormously towards x86's current position.

There is also the question of ROI for consumers. We have half a dozen SunFire servers (circa 2005). In our country this is a lot of money. And guess what, they are lying in stores because there's no modern OS available for them. And the IT will not allow them to operate without security patches and there are no security patches for them from Oracle. And no developer will take up the responsibility of applying a unofficial patch from sources.

And the result? Business line heads will fire anyone asking for non-x86 hardware, unless some customer pays for it.

 

[mwatkins]

I used to work for what was a company in the big systems Unix market. The company doesn't exist any more.

Ironically, they predicted their own demise quite accurately. Despite having roots in RISC, the company made a strategic shift to Intel because, and I quote a founder, "commodity economics always win".

Tom West was right about that but Data General made the transition too late and had too little market share at the time to come out on top (ed: let alone being one of the survivors). EMC bought them, primarily for their storage products.

 

[Simba7]

There is also the question of ROI for consumers. We have half a dozen SunFire servers (circa 2005). In our country this is a lot of money. And guess what, they are lying in stores because there's no modern OS available for them. And the IT will not allow them to operate without security patches and there are no security patches for them from Oracle. And no developer will take up the responsibility of applying a unofficial patch from sources.

And the result? Business line heads will fire anyone asking for non-x86 hardware, unless some customer pays for it.

Why not just stick a modern OS on it? FreeBSD runs great on my old SunFire V100s. Sure, there's that stinkin' 128GB limitation, but you throw a pair of 120GB drives in there with ZFS and RAID0 (or 1) and it runs rather decent.

That's what I ended up doing to mine. I picked up 5 of these for cheap (under $100 shipped) and turned one into a ports server, one into a mail server, and one into a syslog server. I'll eventually find a use for the other 2.

..and if you want to go the Linux route, there's Gentoo. I've tried it on a spare V100 and it runs rather decent. There's still a few bugs in the toolchain, but they're weeding them out.

 

[maxum]

Ms-dos

 

[SR_Ind]

^^

Simba7,

Dumb IT policies. They'd only allow RHEL and Solaris if there is an official version available for that CPU.
Why just Gentoo? If we have the liberty we'd try FreeBSD itself. But doing so, IT will wash its hands off any responsibilities with respect to those boxes. Which means managing patches, backups and downtime all out of scarce developer resources.

 

[rajl]

Just my two cents on preferred linux distributions.

I tried RedHat back in the 7.2 days circa 2001 or 2002. I could never get anything done in it and it drove me nuts.

I found Gentoo in Fall of 2002 and loved it. First, building your system from Stage 1 (several times) really taught me the fundamentals of how the internal system worked. And Gentoo had a rather vanilla implementation, which carried over well to other distros.

I used Gentoo as a primary OS with a backup Windows partition for games and such. I stuck with Gentoo until Ubuntu came out, and switched in May 2005. The things I loved about Ubuntu was out of the box hardware support for almost any hardware you threw at it, as well as the fact that it was essentially Debian, but without associated software being two years behind the times as Debian was back then. Put succintly, Ubuntu was perfect as a desktop Linux for me.

I'm not sure exactly when it happened, but eventually I drifted over to Windows as my primary OS. I don't remember any exact trigger, save that I enjoyed having all my games play on Windows and had no pressing need to use any free oss.

I only recently picked up using FOSS again personally. I first started using FreeBSD for a home file-server. I was afraid to take the plunge, but needed the killer app of ZFS. Turns out that FreeBSD has all the things I love about the Unix family of operating systems, but with none of the things I hate.

I also recently put Linux on my laptop. I moved the Windows license I had on it to an HTPC I built for my wife, and being too cheap to buy another copy of Windows, I decided to see how well Linux worked. At first, I tried Fedora with Gnome Shell to see what all the hoopla about Gnome 3.0 was. I vetoed that choice because even though I was ok with Gnome Shell, I couldn't connect to a wireless network. I then tried Fedora with KDE and had a similar experience. I then decided to go with Ubuntu since I generally have the best results with it and hardware. Their new Unity interface is alright, although it has a few minor behaviors that annoy me to no end. However, installing KDE solved all my usability problems (I guess I really am a KDE guy).

My current vote: Ubuntu. Everything just works as far as hardware goes. If you don't like Unity, installing something else is a breeze. Gentoo is good if you like FreeBSD's ports, although there are some notable differences between portage and ports. If you want rock-solid stable, CentOS and Debian are both good choices although you may lag behind the latest releases of software.

 

[Crivens]

RISC vs CISC is a bit blurred these days.


CPUs are complex enough now that the intel stuff is all RISC internally with a decoder for CISC instructions.

So you get the benefits of CISC style code density but the code is broken up and run as RISC internally. Also, the RISC CPUs got more complex and added streaming instructions like altivec...

And that is one of the problems. The code translation which is being performed eats die area and also power, and it makes the core much more complicated, read, buggy.

I ask why is an SGI or SPARC cpu so inherently superior to x64 these days? Sure in the past they had the memory bandwidth,

Memory bandwidth still is a problem, but it hits them all. And here is one of the benefits which come from CISC, the denser code. It requires fewer memory accesses. On the other hand, the instruction set severely limits the freedom of expression when it comes to planning memory accesses longer ahead so the cache or memory access is not stalling the ALUs. Pipeline length also is a thing not to be ignored.

But it will come full circle. Eventually just like intel ate the server market from the bottom up, ARM is going to eat intel's market from the bottom up as well.

Some time ago I had high hopes in ARM, but not any more.
My next HW will most likely be based on MIPS or some PPC core, but that will be decided when it needs to be.

 

[throAU]

Some time ago I had high hopes in ARM, but not any more.
My next HW will most likely be based on MIPS or some PPC core, but that will be decided when it needs to be.

Eventually ARM will win, imho.

It may take 1 year, it may take 10 years - but the x86 market is eventually going to hit the wall in terms of performance that people actually need in a single die, and ARM is much cheaper, and trailing only a few years now in terms of performance.

Low end servers will start using ARM for better power consumption (and incompatibility with x86 malware), and they will trickle upwards as performance proves to be "good enough" for more of the market.

Same way x86 killed Itanium (PPC, everything else) from the desktop upwards, ARM will kill x86, from the phone/tablet upwards.

 

[drhowarddrfine]

I ask why is an SGI or SPARC cpu so inherently superior to x64 these days? Sure in the past they had the memory bandwidth, etc

Funny you should ask. I was just getting ready to do some painting and that's the only time I wear this shirt anymore (from about 1992 when I worked there).

 

[gore]

I can say why SGI is better with one line:

PC - "Let's make it as cheap and as powerful as possible"

SGI - "F*ck their budget let's make it the best it CAN be, and make sure the Hardware can stand up to being abused worse than Blow...Actually let's make the price about the same as blow!"

PC Hardware seems to fail a LOT more than this stuff ever does. It might cost less for almost the same performance, but it WILL fail faster in my experience. An SGI on the other hand?

Look up their 1600 line Monitors from the mid 90s; They're STILL being sold because they are THAT good.

 

[Simba7]

Look up their 1600 line Monitors from the mid 90s; They're STILL being sold because they are THAT good.

Um.. They didn't make those monitors. Sony did with their Trinitron line.

 

[throAU]

PC Hardware seems to fail a LOT more than this stuff ever does. It might cost less for almost the same performance, but it WILL fail faster in my experience. An SGI on the other hand?

Depends if you buy cheap crappy PC hardware. If you buy proper server quality hardware with redundant PSUs, disks, fans, etc. it is generally fine. I've got PC servers in appalling conditions in the field (stuck in a rack next to a heater in a transportable building on a remote mine site office in Kazakhstan for example - where the temp range outside is -40c to +40c - so the heater gets cranked way up, too) and see a much lower failure rather than "normal" PC hardware. That particular box for example has been in use since 2004 without issue. We have a replacement box on the way right now because it needs upgrading and eventually the disks will no doubt fail.

I've had multiple other PC servers last WAY beyond their intended service life. However, given most busineses depreciate their servers over 4-5 years, lasting much longer than that is somewhat pointless most of the time - it will be outdated and decommissioned by that point anyway.


I'm not saying Sun/SGI have no market. Simply that it will continue to shrink as cheaper-to-purchase x86/x64 boxes become powerful enough for more of the niche higher-end tasks, at far lower cost.

Sure, if your SGI/Sun lasts for 10-15 years it might be worth paying the extra (it might need to last that much longer to pay itself off!), but for most people they want the higher performance that the next generation of hardware offers by that point. If you're a very large company with minimal growth in terms of compute it will probably make sense too. If you're a younger/smaller company with high growth rate, you'll be throwing old hardware out and upgrading anyway.

The cheaper price also means you can more cheaply cluster, keep spares on hand, etc.

 

[Crivens]

Eventually ARM will win, imho.

It may take 1 year, it may take 10 years - but the x86 market is eventually going to hit the wall in terms of performance that people actually need in a single die, and ARM is much cheaper, and trailing only a few years now in terms of performance.

There are other options which are still cheaper and consume still less power.
IMHO, power efficiency will be the driving force for the time comming.
Given the fact that MIPS cores these days cover half the die area an ARM core does, the possibility that ARM is going to win out is anything but sure to me.

Same way x86 killed Itanium (PPC, everything else) from the desktop upwards, ARM will kill x86, from the phone/tablet upwards.

The x86 did not kill anything, people who bought them did, by not buying other things. The question would be why they did so, and superior architectual conceps would not be the reason. The driving force was available software, and that argument is comming to an end, too, as software is available for several architectures these days. Most software, that is.
This is the result of open source like *BSD and Linux, and the switch in CPU architecture which was done by Apple two times showed everybody that it was possible and that there was no real reason to stick to some special architecture.

When I started to use computers, I had no special need like that "MS flight simulator needs to be available" or some office suite. I would have used a SUN, but they were noisy and above all not available in my budget. Same for SGI, but still I did not go the intel road.

Funny you should ask. I was just getting ready to do some painting and that's the only time I wear this shirt anymore (from about 1992 when I worked there).

That, Sir, can be considered sacrilege I still keep my motorola C'Bit gear in good order to remind me of some good time.

And to contribute something to the original thread, the Linux I favor most on my machines is the Linux that is silently sitting on some boot CD and is not installed or running. It was an option as long as they did not try to make it into some fisher price product and as long as they were not under the impression that any file may fit in any directory. But after that, I prefer even Haiku over it (provided it has the SW in need, nothing against Haiku).

 

[gore]

Um.. They didn't make those monitors. Sony did with their Trinitron line.

http://en.wikipedia.org/wiki/SGI_1600SW

That one.

Also, for the other reply about PC Hardware, I think it may have been taken a little wrong, since I do personally use PCs as Servers too, and one in particular has been like amazing!

The very first Computer I ever bought myself, which is an HP Pavilion I bought preinstalled with Windows 98, and that was in like 2000, so that PC, has literally been formatted, partitioned, and repartitioned, and reformatted again, like 200 times, and the little 43 GB HD it came with which I thought was huge at the time, is still going. I also thought 43 GBs was odd. Windows 98 with a Fat File System, would say 42.9 GBs, and Unix based stuff would give it all to me based on File System.

So yea I do get PCs as servers, I mean, heh, it's a FreeBSD forum, which when started, was mainly a PC OS, and LOTS of people used it on Servers.

I always thought that was funny heh; Yahoo! was a huge thing on FreeBSD and it was made mainly for PC Hardware.

Now, one reason I am so into those "REAL Workstations" is the Unix tradition of so much hardware and all of it actually being able to run just one OS. I mean Unix was invented on a machine from DEC, then went to everything. It's amazing if you think about it; This happened at a time when each company made their own OSs and nothing else worked really unless a "real" Hacker wrote another one for it, and yet AT&T was lucky enough to have hired to Legends. True Legends, who changed history forever. That's saying something heh.

 

[Simba7]

http://en.wikipedia.org/wiki/SGI_1600SW

That one.

Ahh.. I stand corrected. That's a rather nice monitor, but it uses an OpenLDI interface. That might make it difficult for it to adapt to recent workstations.

..now if they still made an OpenLDI to DVI (or HDMI) adapter that wouldn't cost an ARM.

 

[gore]

Ahh.. I stand corrected. That's a rather nice monitor, but it uses an OpenLDI interface. That might make it difficult for it to adapt to recent workstations.

..now if they still made an OpenLDI to DVI (or HDMI) adapter that wouldn't cost an ARM.

Heh, yea. Those Adapters seem to range quite a bit. But yea that Monitor, is just amazing. I've always Loved SGI stuff though. The first time I saw an SGI Computer, I was like "OMG! They make something other than a white or gray box!" Because it was like Neon Purple or something. I've just always Loved the whole case design thing they went with.

Once I've finished paying off everything I've got going right now, I do plan on buying an SGI Workstation and everything. I know they're dated and all, but I Love them, and I think they truly are works of art. So it doesn't hurt if I get myself a Workstation running IRIX, and play Quake2 on it

 

[roddierod]

Once I've finished paying off everything I've got going right now, I do plan on buying an SGI Workstation and everything. I know they're dated and all, but I Love them, and I think they truly are works of art. So it doesn't hurt if I get myself a Workstation running IRIX, and play Quake2 on it

You should keep an eye on Craiglists, local FreeCycle groups or even local LUG mailing lists. I just got 2 SGI O2s, a SGI Fuel, 3 copies of IRIX, 3 21" SGI CRTs and other misc stuff for free by answering a posting on one of these.

 

[drhowarddrfine]

I just got 2 SGI O2s, a SGI Fuel, 3 copies of IRIX, 3 21" SGI CRTs and other misc stuff for free by answering a posting on one of these.

When I first went to work at SGI, I called my boss to ask what kind of system he was getting me. I was hoping for a 4-processor system. "Four?!", he said. "I was putting you down for a 16-processor system."

I was the only one working out of the St. Louis office. I was in heaven.