I’ve got quite a few comments on my post on Debian, which is always nice. But I now feel the need to more or less defend myself. To set some things straight first off, I was using Debian etch x86_64. I did a clean install using as many “default” options as possible. Which is the same route I took with openSUSE and Wolvix as well. So that, along with the reminder my hardware never changed, I made several subjective observations.
The observation that got me the most flack, as one would expect, was that I found Debian bloated and slow. In the end I did not attempt to speed Debian back up. Can Debian be fast? I’m sure with enough tweaking any distribution can be fast. But out of the box Debian doesn’t feel fast to me. Again a subjective claim on my part. Like or not I don’t like Debian. I’m sure the install can be tweaked and the system post-install can be tweaked. Well thats great but that wasn’t what I was looking for.
Finally, my thoughts as to why all of this is. Well, again subjectively speaking here, I believe it a lot of comes down to compile options. Its not secret that distributions that are, within the Linux community, considered “slow” (i.e. Ubuntu) have their packages compiled with as many flags as possible. In the Gentoo community we’ve renamed them USE flags but its the same concept. Other binary distributions that are generally considered “fast” (i.e. Arch, Slackware) tend to take a “Gentoo approach” to compilation choosing to leave out a lot flags that their developers consider unneeded. And if these flags are needed by the end user it is generally left up to the end user to recompile them on their own. Thus you end up with tools like the ABS (Arch Build System). This combined with a “necessity only” attitude toward packages included by “default” your left with a clean, healthy, and generally quick system. Can I prove any of that objectively? No. And I’m not going to try. Its just something I’ve observed during my short stay here within the Linux community and I know a lot of people who would disagree with me.
Either way I really don’t care to be honest. The buck stops here as far as I’m concerned. I’ve said my piece and thats all there really is to it.
Enjoy the Penguins!
I don’t really understand what your problem is. Debian is not slow. Actually Debian is one of the few distributions which runs wery well on older hardware too. You might not have noticed ot but Debian is available even for the m68k based Amigas which are not really fastest computers on the planet anymore. Debian is used on many embedded and mobile devices like Nokia N800. Those mobile devices are not speed monsters but Debian runs just fine on those hardware too. Debian runs much better on older PCs too than most of the other distros. Eg. Ubuntu is wery slow on 300Mhz Celeron with 128MB ram, Debian runs much better. Ofcourse that hardware is not blazing fast no matter what distro you run. But anyway Debian sure is not slow.
Actually Nokia N800 InternetTablet OS 2007 is just based on Debian. But anyway what was said applies to it too.
Why did you remove this comment? You don’t want people to see it? Sorry but this is the truth:
I don’t really understand what your problem is. Debian is not slow. Actually Debian is one of the few distributions which runs wery well on older hardware too. You might not have noticed ot but Debian is available even for the m68k based Amigas which are not really fastest computers on the planet anymore. Debian is used on many embedded and mobile devices like Nokia N800. Those mobile devices are not speed monsters but Debian runs just fine on those hardware too. Debian runs much better on older PCs too than most of the other distros. Eg. Ubuntu is wery slow on 300Mhz Celeron with 128MB ram, Debian runs much better. Ofcourse that hardware is not blazing fast no matter what distro you run. But anyway Debian sure is not slow.
You’ve summed up the problem of distros like Ubuntu and Debian pretty perfectly: They wanna offer one CD/DVD/package that fits all. All X86 packages for Debian/Ubuntu are the same regardless of what hardware they might be running on.
That is not “better” or “worse” than the Gentoo approach, it’s merely different with a different focus. You get rid of some of those Gentoo specific problems (like some people using insane USE-Flag/CFLAGS combinations) but you pay that with less speed and more “bloat”.
Just to clarify, I haven’t removed any comments. The only ones I get that don’t make it are spam or include vulgar language. Other than what you see is all I’ve got.
I’m still waiting for the first benchmark that shows that compile options have any significant effect at all. There is a handfull of packages (such as the kernel and ssl) that really benefit from a customized compilation, but in my experience it doesn’t matter at all for most packages.
Lol, the reactions are funny. Lighten up everyone.
@Steven
In the thread I mentioned earlier Paul posted the suggestion that a manually compiled kernel can be the main reason why Gentoo feels faster then other distro’s.
@Miksuh
I have ran Debian on old hardware and it felt dog slow in comparison to Slackware and Gentoo.
Did you know what the main complaint was of the N800 and it’s predecessor? That is was slow!
I wonder have you used Gentoo yourself?
Concerning Speed of Gentoo and compile flags.
Its very much a mistake to think that the CFLAGS make a huge difference. Once you’ve got O2, the rest is wasted.
What (can) make gentoo fast is the choice of libraries to link against.
If you have generic distribution it going to link against myql, maybe openldap even kerberos.
And at startup, the loader for the apps is going to resolve those dependencies and load up those libraries.
That, my friends, is slow. Its so slow that the technique of “prelink” was invented to ameliorate it.
Its so slow that the gnome developers are trying to pull the subsidiary libraries into gtk+.
I’m sure that if Debian was used on a Nokia, then it would be target compiled for the environment–No kerberos–!
Regards,
John
Your probably right John Huttley, but I never said anything about CFLAGS.
I’ve read a few discussions on CFLAGS. Most don’t make a difference except in particular circumstances (leaving them to be best-maintained by package maintainers or upstream). However, depending on the arch they can have a significant effect.
I’d be interested in seeing some benchmarking done on the use of system-wide CFLAGS. The more widespread the use of a compiler-option is the large the impact is likely to be. Personally, I’ve tended to stick with -Os – most systems are pretty limited in RAM and benefit from reduced memory use. The increased cache hits and decreased swapping should more than make up for a few cycles wasted in a loop. However, I wouldn’t expect any benefit to show up if just one package were compiled with -Os – memory use is accumulated across all running software so the benefits probably tend to add up.
However, I do agree that stripping out unneeded dependencies can have a much bigger impact. It also has a potential impact on code size and memory use.