On Tue, Jul 19, 2011 at 10:49 AM, Arun Khan <[email protected]> wrote: > On Tue, Jul 19, 2011 at 7:14 AM, Amarendra Godbole > <[email protected]> wrote: >> What good is efficiency when it sacrifices stability and reliability? > > I find the statement to be contradictory - suggesting that Linux is > efficient but not stable or reliable. > IMO, all three have to co-exist.
Let me make it clear at the outset - Linux here refers to the kernel, and the GNU userland, and not just the kernel, since both are required for a stable, reliable and robust system. "Have to" may not necessarily mean "does". ;-) I don't trust code that has a feature-bloat, and tries to support everything under the sun, including all possible languages known to mankind (for eg., why do you think setlocale() exists at the start in every single GNU userland utility?). Overall the S/N ratio for anything in linux is very low due to the additional baggage it carries. Additionally, large corporations who contribute people to the kernel may have their say (which is why they have the people there in the first place). >> I find Linux code to be bloated, untested, and added at whim of a >> developer or two who has/have clout. All Linux h/w drivers are >> typically "blobs" - written by the device manufacturer, and used >> "as-is". I doubt if a serious scrutiny of these happens, since Linux >> requires to support all known h/w to mankind in its race to world >> domination. > > Please quantify your statement about "All Linux h/w drivers are > typically "blobs" - written by the device manufacturer," http://www.emulex.com/downloads/emulex/cnas-and-hbas/drivers/linux/736-driver-and-hbanyware-kits.html came up with a google search. You get a binary rpm with install and un-install options. Emulex mostly wrote the source, and I don't know if its available (I guess not - since it will give away their secrets). Many hardware vendors choose to keep their specs secret, and make a binary driver available for linux. Your choice is to use it "as-is". This is how Linux derives it huge support for hardware - vendors keep specs secret, and give out binary blobs, and also commit to upgrade those as the versions of the kernel progress. They get business, and a over-inflated ego of supporting open source. Also look at http://damien.bergamini.free.fr/packages/openbsd/iwn-firmware-5.6.tgz where the wireless driver firmware for Intel WiFi Link 4965/5000/1000/6000 IEEE 802.11a/g/n wireless network device is available. Intel does not make specs available, so if I have to have this wireless card supported in OpenBSD, I must use these blobs which come straight from intel. I do not have the source code for these, and I am sure Linux uses the same blobs. I have no idea about NetBSD or FreeBSD, but OpenBSD moved away from this blob-culture, simply because it was too difficult to trust these binary blobs. As a result the supported h/w list is narrow. Vendors who are willing to open their h/w specs do get a driver written by the OpenBSD developers. > Wow, if this were true then users would not be struggling to install > or run Linux with their cutting edge hardware or get their USB dongle > to get detected or work with xyz service etc. One could simply press > F6 and plop in the "manufacturer's" driver CD. I'd rate OpenBSD generic driver support far better than Linux's from my experience. > The development model of the OSs in discussion are different. I > either case end users have the "freedom" to choose. About a couple > of years ago, I seriously considered FreeBSD as a NFS server for a > client (NFS server being better on FreeBSD). The server had a > Chelsio 10Gb (2 port) NIC. After doing some search I found that > FreeBSD supported the card but users had reported problems with the > driver (stability). I opted for CentOS 5 that has native support > (FOSS driver) for the card. http://service.chelsio.com/drivers/ All drivers contributed by Chelsio - so they decide whom should they work with (compat). Also note the copyright rests with Chelsio, so they can at their will discontinue the support any time. > I use various flavors of Linux distributions (which amongst themselves > have their own pros/cons) as well as FreeBSD (to a lesser extent). > My latest discovery is DragonFlyBSD and plan to give it a spin. [...] Understand that I am not saying Linux is bad -- if you have no choice, you have no choice, as you example in the Chelsio case above demonstrates. If I were you, and if this were a customer case, I'd have done the same thing. It, however, does not make Linux more stable, or reliable, or robust. The smartness lies in "using the right tool for the job" - OpenBSD is right for me so long, and I plan to continue using it. It also contains a lot less philosophical baggage than Linux. ;-) Have fun with whatever OS you choose - and that's the bottom line. -Amarendra _______________________________________ Pune GNU/Linux Users Group Mailing List
