Bloatware is not about size, but about inability to see though the system.
From beginning, there were distributions which had their package management system adapted to modification and as such granting easy transparent changes in system;
and there were systems which simply, like pure LFS, drop binaries on top until mess happens, very similar to winblows.
I have been testing RR and later sabayon several times, and each time I failed to understand any pro's of this distro.
A very similar distros to this were - dreamlinux and foresight, some base taken, pieces thrown over and the heap starts falling apart - its only matter of time.
Gentoo. However I approached gentoo separately and much later after first contact with RR.
I plan to move to calculate linux, because it does many things right - it is collective effort to push desktop and small business gentoo variations.
The difference? You take gentoo and modify it - destination desktop and server (install dvd and fast system unroll included).
But you write every change in ebuilds, if it is improvement - to gentoo directly, if it is own non-cannon modification in own overlay.
Then, the system is really gentoo (but respin), really uses gentoo portage and really 100% gentoo compatible.
It is like buying ready coffee drinks instead of growing on own plantage, with only this choice being difference, not compatibility.
More than that, the overlay has own profiles and is also visible from official gentoo tree (layman).
Imagine you can do stage3 gentoo install and immediately switch to calculate profile within gentoo itself and "upgrade" your gentoo to calculate. Or backwards.
You can cut and paste system with it just as you do with gentoo.
This is called gentoo-based distribution.
Binary packaging has been weak side of gentoo for a reason... this is a source distribution for sake of easy modification and customisation.
However not everyone wants to customize every ebuild... mostly its use flags and sometimes cflags help accelerate a bit.
The code itself often uses inline optimizations, making recompile nearly useless and sometimes dependencies are not that hard (see apt/synaptic "recommends" features).
So using binary packages, where it makes sense to use them(like cutting down installation time) is really a good addition.
For this matter, gentoo own portage has whole set of tools to generate binary package and install it. Portage tree also includes binary packages and tracks their state hard so they do not segfault over time. But this is only done for reason of 1)popular and HUGE package without any useflags that make much sense 2)binary-only package
Now, "sulfur" and "entropy" apart from being very strange, hard to associate, words (adding complexity on top of existing portage system) are no way even near apt-get.
Seriously, waiting for sulfur to update the package tree in FIVE minutes, watching Ads at same time, without ability to use ANYTHING gentoo or features similar to apt (recommends, optional dependencies). Sulfur and entropy are not improvement of gentoo, they are brain-replacement, they conflict with portage. Why not to take apt-get instead? Why not generate ebuild-to-dpkg script and make dpkg gentoo repo?
Apt is state of art, the speed and features which it has are simply overwhelming. But you can't use it with portage too, because it is binary fixed-version dependency-tracking system with possibility to pull source code as well, but never to use source code instead or mixed with binary.
The sabayon idea of having already compiled tree conflicts ITSELF with gentoo. It is no way improving gentoo, NOR it is gentoo.
Gentoo ofers multiple versions - entropy castrates that and many more.
Only preselected unchangable versions, hard dependencies, inability to use anything from what makes gentoo - gentoo(use flags, profiles etc).
Its just huge and slow blob of blobs, sitting on top of distribution and blocking gentoo from being gentoo. And it is not polished binary package manager, why not take already polished ones instead of inventing the bycicle?
The binary package manager that gentoo could really profit of, is completely different from anything that exists.
I guess it would be nearly completely autonomic, decentric p2p-hosted, yet VERY firmly synced to official portage tree, system, that could be installed on any (user) machine, with administrator providing it with cache-space, network bandwith and ability to offload locally built packages to that p2p tree(just like torrent, but automated).
The system that will hook into portage and on every emerge (if enabled to do so), lookup if already built package with completely same flags exists on the network.
Then, judging from package size, priority (maybe user does not want to build, at all), the system would decide - to launch emerge, or to pull the existing package.
THAT would be improvement.
Exactly. This means, sabayon package manager and gentoo package manager - conflict.
This speaks of "quality" of built packages.
If it is required to pull a package once, that would make portage functional, BUT this package had already different version (I guess more latest and incompatible with portage) on existing system, means they either do not use portage to build all packages, or they mess up everything. Strange case.
I do not use calculate linux, but I have talked with its main developer before and he commented that sabayon has very strange habbit of putting everything in kernel and every driver in initramfs.
Calculate kernel generator, on contrary, does live-check of present hardware and puts only that into initramfs. The rest resides as normal modules on root partition(if installed). Not only this approach is more efficient, it also much more sane.
The only difference between gentoo and calculate are profiles (prefered packages), some own ebuilds(which are available on portage), as well as fact that most of the stuff is delivered in prebuilt state. But not in blob-like mess.
Sorry, if I'm wrong somewhere, in fact I would really really appreciate if you tell me your opinion.



Reply With Quote
