Weekly Edition
Return to the Press page| Weekly edition | Kernel | Security | Distributions | Search |
| Archives | Calendar | Subscribe | Write for LWN | LWN.net FAQ |
Sun Microsystems' Next Linux Move (Seeking Alpha)
(Log in to post comments)
MySQL = credibility?
Posted Apr 3, 2008 18:03 UTC (Thu) by ncm (subscriber, #165) [Link]
Did buying MySQL really give Sun credibility, instant or otherwise? It's hard to guess what they thought they were getting. Two possibilities... First, they might have customers who would like to use MySQL in some way that MySQL AB wasn't prepared to license, but Sun would. Second, they want to steer development in some direction that M.AB wouldn't have wanted to go. It's hard to see the first possibility being worth $1B, unless it involved a government contract. The second seems more likely. Sun has said it's committing to massive parallelism, so maybe they want to parallelize MySQL to make theirs the preferred high-end database engine.
MySQL = credibility?
Posted Apr 3, 2008 18:31 UTC (Thu) by epa (subscriber, #39769) [Link]
You could write an SQL-based relational database from scratch for a lot less than a billion dollars, and it would be better than MySQL.
MySQL = credibility?
Posted Apr 3, 2008 18:42 UTC (Thu) by Klavs (subscriber, #10563) [Link]
AFAIK MySQL does turn a profit - so atleast they're not buying a sinking ship. if they choose to grab the wheel and steer it, so that it sinks is just something we'll have to wait and see :)
MySQL = credibility?
Posted Apr 3, 2008 19:09 UTC (Thu) by tzafrir (subscriber, #11501) [Link]
http://www.postgresql.org/about/sponsors http://www.postgresql.org/community/contributors/ Sun contributes to pgsql and employs one of its core developers. I'm not going to claim that pgsql is better than mysql or vice-versa. But it certainly does not require building from scratch.
it's not that simple
Posted Apr 3, 2008 21:18 UTC (Thu) by tialaramex (subscriber, #21167) [Link]
• Writing an RDBMS is not something that can be done overnight, for any money. MySQL exists now. For a billion dollars you get a new RDBMS in 2010 or beyond. Meanwhile your strategy which needed MySQL sits dormant and maybe becomes irrelevant. • Projects fail. Short of spending $100 million on each of ten different projects to build such an RDBMS you can't have any assurance you'll get a product from this approach in reasonable time at all. Sometimes for any number of political, social or technical reasons a project loses traction and has to be abandoned at great expense. This is not a problem to which rich corporations are immune. Microsoft had to "reboot" the Vista project to get a releasable operating system, and Apple had to abandon Copland altogether and just buy in a replacement. • You still don't have MySQL but only some new RDBMS. If your strategy requires MySQL this is no help. Even if your strategy only requires a drop-in replacement for MySQL, this may cost an order of magnitude more than you expected, simply because bug-for-bug compatibility is really tremendously hard. A single unexpected feature can send you back to the drawing board, throwing away months of work, "Aha, their internal data structure for this type of information isn't a tree after all, so our design produces output in a different order".
True
Posted Apr 3, 2008 22:30 UTC (Thu) by man_ls (subscriber, #15091) [Link]
Plus, MySQL is the 'M' in LAMP. After building a fast and robust database, even if it was 100% compatible, how long would it take to have a sizable user base? What about a development community?
it's not that simple
Posted Apr 3, 2008 22:40 UTC (Thu) by epa (subscriber, #39769) [Link]
Yes, the hard part is not writing an SQL database system, it´s making one bug-compatible with MySQL. But why would Sun want that? Which customers are particularly dependent on MySQL and no other database? The software that needs MySQL is usually low-rent stuff, online bulletin boards like Slashcode. It´s unlikely any banking applications, for example, are using it. Yet these are the high-paying customers.
it's not that simple
Posted Apr 3, 2008 23:46 UTC (Thu) by chromatic (subscriber, #26207) [Link]
A lot of people see MySQL as an excellent example of the Innovator's Dilemma. Ignore the low-rent customers at your own risk.
it's not that simple
Posted Apr 4, 2008 3:25 UTC (Fri) by drag (subscriber, #31333) [Link]
I don't know why people like to shit all over Mysql. It does it's job well and provides the levels of features that it's users desire. It's a backing store for most all open source online applications. That's all. That's all databases are for most people, just backing store. Then on top of that they have all sorts of different back-ends that provide any sort of features that you want if you need them while not requiring users to change very much in their apps to take advantage of it. This gives it a level of flexibility that I don't think your going to find in any other SQL-ish database. It _works_. And it's not just used by small stuff either. I am pretty sure that people have successfully used Mysql in larger and more important sites then anybody has ever done with postgresql. I mean it's not like postgresql is the paragon of correctness, stability, or anything else that is important in databases. I am sure that it's fine, wonderful, and great, but why all the hate? Is everybody that likes postgre feels like sand has been kicked in their face by the big bad commercial Mysql bully or something?
it's not that simple
Posted Apr 4, 2008 3:27 UTC (Fri) by drag (subscriber, #31333) [Link]
No offense, but I am just amazed that a any sort of mention of Mysql in any sort of way in any sort of article there are always a certain number of people that just love to talk about how much it sucks compared to everything else.
it's not that simple
Posted Apr 4, 2008 8:34 UTC (Fri) by nix (subscriber, #2304) [Link]
Oh yes. SQLITE IS MORE FEATUREFUL RAH RAH RAH (er. perhaps not. God knows mysql is more *usable* than bloody Oracle.)
it's not that simple
Posted Apr 7, 2008 8:47 UTC (Mon) by liljencrantz (subscriber, #28458) [Link]
Here, here. Anyone who claims MySQL is a toy should take a long hard look at Oracle. * You cant use column aliases in the having clause, e.g. "select foo as bar ... having bar>0" is a syntax error. * Null is the same as an empty string. * 30 character maximum on table and column names. * Varchars can't be longer than 4000 characters, for anything longer, you have to use clobs, which are glacially slow. * The Oracle character set named utf-8 is not actually utf-8. The character set name al32utf-8 is actually utf-8. * Passwords are case insensitive. * Oracle uses two different but very similar languages, SQL to perform simple queries and PL/SQL to define functions, procedures, etc. These languages have different definitions of the basic types, e.g. the maximum length of a varchar variable is different in the two, one has a boolean type, the other does not. * The oracle installation is loads of fun. At one point in the installation, a popup window appears prompting you to something like "open a new terminal as root and run the following commands: ...". You also manually have to edit about half a dozen system files. I don't know how a program can have those kinds of design issues and still be considered even remotely enterprise ready. Sure, MySQL has various limits on how many triggers you can define, but that can be worked around pretty easily. By default, it is far to forgiving about bad data, but that is a configuration issue. It is surprisingly easy to corrupt the database on crashes, etc.. but that is true for Oracle as well and in both cases it is almost always the case that the logs can be used to restore the database. MySQL is glacially slow on subselects, but there are loads of other types of queries where Oracle is dog slow and if you write your SQL with MySQL in mind, your queries will usually fly. Remind me why Oracle is better than MySQL?
it's not that simple
Posted Apr 7, 2008 14:52 UTC (Mon) by nix (subscriber, #2304) [Link]
* Null is the same as an empty string.Now, now, don't go casting false aspersions. They changed this, silently: NULL is now the same as an empty string *in WHERE and HAVING clauses only* except that if the optimizer chooses to do particular things to your query or your database has particular patches, it isn't.
Isn't that clearer?
Remember also its lovely flow analysis in the vile abomination that is called Pro*C: pass an uninitialized char array into the *OUT parameter* of a PL/SQL function, or into a SQL statement's INTO clause, and it does no flow analysis at all and strlen()s the uninitialized string prior to sending it to the server. Great design, guys.
(I'm not even getting into Pro*C's other manifold faults because this post would be 300K long.)
* Varchars can't be longer than 4000 characters, for anything longer, you have to use clobs, which are glacially slow.And PL/SQL can only manipulate clobs by cutting them into pieces (whose maximum length is the same as the maximum length of a VARCHAR, which is not 4000 characters in all situations but rather depends on your *database character encoding*) and manipulating them that way!
* Oracle uses two different but very similar languages, SQL to perform simple queries and PL/SQL to define functions, procedures, etc. These languages have different definitions of the basic types, e.g. the maximum length of a varchar variable is different in the two, one has a boolean type, the other does not.It's worse than that. There are separate implementations of PL/SQL in each of Oracle's products. These are different languages with distinct bugs, different 'specifications' (most nowhere documented that I can find) and sometimes-diverging development histories. There is no way to find out which language/VM you are running against at compile- or runtime. Many of the bugs are catastrophic: e.g. Oracle Forms 9's PL/SQL implementation allegedly supports the CLOB and TIMESTAMP types, but if you use it the Forms frontend sprays incomprehensible errors at you (in the latter case) or instantly terminates (in the former).
Remind me why Oracle is better than MySQL?I'm a PostgreSQL man (just never used MySQL much and heard enough about it to stick with what I like). It has faults (the lack of full support for table inheritance for one) but like it or not at least they try for consistency, which is something that's wholly lacking from the big database vendors.
it's not that simple
Posted Apr 7, 2008 17:23 UTC (Mon) by liljencrantz (subscriber, #28458) [Link]
Wow. I knew about the clob limitations, but the rest... Had no clue. I think it's awesome that Oracle is the 9:th largest contributor to Linux. I'm sure they'll make Linux enterprise ready in no time.
it's not that simple
Posted Apr 7, 2008 18:14 UTC (Mon) by nix (subscriber, #2304) [Link]
i.e. bugger it up completely? Yep, that's possible, but I hope not! I've been doing Oracle stuff for ten years now and much of it consists of spending great effort to find ways to *avoid* using Oracle, because using it is so unpleasant. Using Unix stuff instead is generally vastly more enjoyable. I suspect a lot of the ugliness in the Oracle stuff is because their competition is largely other hugely expensive databases like SQL Server, which are *also* incredibly wart-filled, because they're all based on really really old code and have been maintained for decades by people who care more about getting the next version out and getting a new marketing buzzword implemented than they do about internal consistency. (e.g. Oracle's 'object oriented' features introduced in Oracle 8. There are *still* bugs in there whereby unprivileged users can crash entire database instances. They *still* don't implement inheritance as far as I can tell. It's *still* nearly useless for any practical purpose... because it was never intended to be *used*: it was just a marketing buzzword, when 'object oriented' was the latest big thing.)
it's not that simple
Posted Apr 4, 2008 13:38 UTC (Fri) by i3839 (subscriber, #31386) [Link]
All I'll say is that for teaching SQL to students, MySQL is rather unfortunate in its limited support for standard SQL features. That it requires workarounds to let it accept otherwise fine SQL code only make it more annoying. That MySQL isn't alwasy smart enough to optimise more complex queries and thus forces you to write ugly queries instead, isn't very nice either. Let's say that from the people who have experience with MySQL and other databases I also don't hear great things about it either, to put it mildly, so it's not only my own unfortunate experience. But I'm sure the newest version is better.
it's not that simple
Posted Apr 7, 2008 13:06 UTC (Mon) by liljencrantz (subscriber, #28458) [Link]
All databases have their warts. Postgres arguably has fewer warts than MySQL, but it's not perfect either. For one thing, table creation and simple insert statements are often orders of magnitude slower than in MySQL, in my experience. This is using the latest Postgres version in Ubuntu dapper. Also, having to vacuum the tables to reclaim storage is a bit silly.
it's not that simple
Posted Apr 7, 2008 15:00 UTC (Mon) by nix (subscriber, #2304) [Link]
PostgreSQL table creation is a bit slower because it's transactioned :))) personally I'm willing to accept a bit of sloth to gain the benefits of being able to roll back on screwups. With transactioning, you can easily have any random thing create tables and vape them automatically on ROLLBACK: something which should have been the case in all relational DBs from the very start, I think. Vacuuming has been automatic for about a year now.
it's not that simple
Posted Apr 7, 2008 17:39 UTC (Mon) by liljencrantz (subscriber, #28458) [Link]
I don't know about table creation, but MySQL inserts a lot faster even when using transactions and InnoDB, at least in my experience. And vacuuming is still silly, even if it's done automatically. Last time I checked, it has a noticable performance impact on any queries running while vacuuming, meaning Postgres is not very good at handling continously high loads. I like Postgres just fine, don't mean to slam it at all, every database product has some warts. MySQL probably has bigger warts, but both are in my experience pretty nice systems to deal with, unlike say... Oracle.
it's not that simple
Posted Apr 7, 2008 18:16 UTC (Mon) by nix (subscriber, #2304) [Link]
Yeah. Actually as far as I can tell you have two choices if you want transaction consistency: use multiversion concurrency control, or use locks. Locks have a *much* worse performance hit in the concurrent case, and annoy developers more, but a lower hit in the single-threaded case. It's a toss-up which to use :) (Oracle moved to MVCC a while back, too, while keeping around a huge pile of grot to try to mollify people who were used to locking everything and messing around in horrible ways with 'rollback segments' and that crud.)
it's not that simple
Posted Apr 4, 2008 5:49 UTC (Fri) by AJWM (subscriber, #15888) [Link]
> I mean it's not like postgresql is the paragon of correctness, stability, or anything else that is important in databases. Some years back when I was choosing an RDBMS for a project I went with PostgreSQL for a couple of reasons. At the time it was certainly more correct, stable and so on than MySQL -- that may well have changed as MySQL has improved. The other important criterion was that it supported Embedded SQL in a manner almost identical to Oracle (one or two trivial #ifdef differences, much closer than any other Embedded SQL I'd seen). I don't think MySQL does yet, does it? But sure, if all you're using the DB for is a backing store, there are plenty of options. Actually I think if PostgreSQL had had a name as pronounceable as MySQL, it would have become the clear front runner a long time ago ;-)
Sure, see you in 10 years
Posted Apr 3, 2008 21:26 UTC (Thu) by coriordan (guest, #7544) [Link]
Maybe they could write a better database for less, but if they want one as debugged as MySQL, it's gonna take 5-15 years. It's not about making something exist, it's about having it now.
MySQL = credibility?
Posted Apr 3, 2008 18:56 UTC (Thu) by dmarti (subscriber, #11625) [Link]
Or they could just have bought it for the same reason they bought StarOffice. Commoditize the Complement.(And besides, if you run a proprietary software company, a dollar in another proprietary software company's bank account is a bad thing. The other company will just spend it gaming the legal system, offering low up-front prices and locking in customers, and bidding up programmers' salaries. The poorer and more desperate Sun is, the better Oracle does, and vice versa.)
MySQL = credibility?
Posted Apr 3, 2008 22:20 UTC (Thu) by ncm (subscriber, #165) [Link]
You seem to be saying that this gives MySQL credibility, more than it does Sun, and at Oracle's expense, leaving more money for the customer to spend at Sun. I'll buy that, but not for $1B. It might be a defensive move, if there was a threat somebody else (IBM? HP?) might buy it instead.
add it up
Posted Apr 4, 2008 1:09 UTC (Fri) by dmarti (subscriber, #11625) [Link]
Yeah, it's a few hundred million of MySQL's value as a going concern, a few hundred million of anticipated value in being able to sell it to Sun customers, a few hundred million of commoditizing the complement, and a few hundred million of Valley "ha ha u r l@merz!!!"On the other side, we get Oracle supporting btrfs, so yay Freedom.
MySQL = credibility?
Posted Apr 4, 2008 0:02 UTC (Fri) by jordanb (subscriber, #45668) [Link]
MySQL AB was a pretty small company with a pretty small name. Their product was widely deployed but not in situations that really bring in the big contracts. I suspect that Sun figures they could buy up the MSQL product and existing customer base, and then leverage its own presence to get MySQL into places where the really lucrative contracts live.
Sun Microsystems' Next Linux Move (Seeking Alpha)
Posted Apr 3, 2008 19:30 UTC (Thu) by lmb (subscriber, #39048) [Link]
That obviously makes perfect sense for Sun. They are first and foremost a hardware vendor; they want to marginalize the "independent" OS vendors as much as possible and eat into their margins. (Just like Oracle as an application vendor tries to marginalize everything below their apps.) On the other hand, it doesn't mean that Red Hat or Novell lose any business; if the effect of Sun's action is growing the market, this can be beneficial to everyone involved.
Divide and rule
Posted Apr 3, 2008 19:31 UTC (Thu) by cyperpunks (subscriber, #39406) [Link]
This is just Divide and rule strategy from Sun.
They call it GNU/Linux - thanks Sun!
Posted Apr 3, 2008 21:31 UTC (Thu) by coriordan (guest, #7544) [Link]
One move they made was that they've started calling it GNU/Linux:
Not all the time, but half the time, and that's a big improvement. Stallman put online a good article not so long ago:
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 7:36 UTC (Fri) by michaeljt (subscriber, #39183) [Link]
The policy of calling the whole system "GNU" does beg the question - what proportion of the average free software-based system is actually GNU software? And of that software, how much of it can be considered part of the core of the system and how much as tools? Does the compiler suite count as a development tool or as part of the system core? Does the fact that it is required to build the system affect the equation? Does Gnome consider itself to be primarily part of GNU (is Solaris GNU?), or an independent project? Which proportion of free software systems (desktops or otherwise) use Gnome? And while we are at it, what proportion of such a system is actually Linux? (As a side note, I like to talk about the systems as being "Debian", "Ubuntu", "Gentoo" or whatever in order to sidestep the issue. That at least seems to me to be an uncontroversial description of them.)
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 7:58 UTC (Fri) by coriordan (guest, #7544) [Link]
David Wheeler looked into this in his paper More than a Gigabuck, saying:
...the total of the GNU project's code is much larger than the Linux kernel's size. Thus, by comparing the total contributed effort, it's certainly justifiable to call the entire system ``GNU/Linux''...
And he didn't include GNOME when doing that count. As the system gets bigger, the GNU proportion may decrease, but it was still 8 years of work by the GNU project that made most of an operating system before Linus's 1991 kernel project.
Most large companies push the name "Linux", and push the myth that Linus Torvalds developed an operating system in 1991. It's an easy push since "Linux" is a shorter name than "GNU/Linux" and because journalists love printing the story of the college student who spent the cold months of one year writing an operating system that's competing with that of the world's richest man. It doesn't matter that it's not true, it's a great story and no one will check :-/
The reason these companies push that myth is because they don't want their customers to hear the GNU philosophy and start thinking that they deserve certain freedoms from all the software they use. That would hurt profits.
I'm glad Sun is bucking the trend.
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 8:20 UTC (Fri) by michaeljt (subscriber, #39183) [Link]
Interesting paper. So by far the largest part of GNU's contribution to Redhat (we can probably generalise from that) is the development suite, in which I would include Emacs, and glibc. I would not count glibc as "just" part of the development suite, as an "average" computer user does not need a compiler and friends on their computer, however indispensable they may be to create the system, but try to remove glibc from a GNU/Linux-based system... Re the Linux myth companies though, my personal suspicion that they just don't care much about the GNU philosophy rather than actively trying to sabotage it. Never attribute to malice what can be explained by laziness :)
They call it GNU/Linux - thanks Sun!
Posted Apr 5, 2008 19:22 UTC (Sat) by rsidd (subscriber, #2582) [Link]
If the toolchain (development tools) were a reason to call the project GNU/whatever, then "GNU/" would need to be prefixed to the BSDs and Mac OS X too. Even RMS does not make that claim.
As for glibc, the maintainer of many years, Ulrich Drepper, made his opinion clear some years ago.
Many other userland utilities and libraries on linux systems, that would be regarded as essential features of a Unix-like system, come from BSD and other sources, not GNU.
RMS's claim seems to be that he had a vision of developing a free operating system, using pre-existing software as far as possible (notably X and TeX) and filling in any gaps (notably toolchain, C library, coreutils) via the GNU project; though the biggest gap of all, the kernel, remained unfilled in 1991, Linux came along and made use of GNU work (just as GNU had made use of MIT's or Knuth's work); therefore the combined system must be called GNU/Linux, even though GNU never felt the need to acknowledge other projects in this way.
Moreover, his claim minimises the importance and difficulty of the kernel. The GNU project has failed to produce a usable kernel, not only in 1991, but to this day. And the main reason Linux has more users than the BSDs is the kernel: most software runs fine on BSDs, but their kernels lag in hardware support and modern features. Nobody has written a free kernel comparable to Linux from scratch; but there are several free C libraries, a few free compilers (eg TenDRA, LLVM), and BSD versions of almost all the GNU coreutils and fileutils.
So insisting on GNU/Linux sounds very much like stealing credit. RMS had the big ideas early on, and he is credited with them. He deserves no direct credit for today's working Linux systems -- at least, no more credit than hundreds of other contributors.
It's more than GCC
Posted Apr 7, 2008 9:49 UTC (Mon) by coriordan (guest, #7544) [Link]
There's Glibc, which cannot be replaced (except for toy projects and purpose built tiny distros).
There's Bash, coreutils, diffutils, findutils, grep, gzip, readline, tar, screen, etc. Most of these could in theory be replaced by BSD software, but the easiest way to do that is to wipe your GNU/Linux installation and install FreeBSD :-)
The "it's replaceable" argument actually implies removing the "Linux" from "GNU/Linux" since GNU cannot be replaced (glibc at the very least) but the Linux kernel can and has been replaced with the kernels of FreeBSD, NetBSD, and Solaris.
Writing development tools was an essential part of making an operating system. In fact, some of GNU's best work was in enabling others to help by producing the development tools and the licences. As well as GCC and GDB, the GNU project wrote Make, Gawk, GNU sed, Bison, binutils, gettext, among others.
We're not talking about a one trick pony here.
It's more than GCC
Posted Apr 7, 2008 14:55 UTC (Mon) by nix (subscriber, #2304) [Link]
glibc is replaceable by uClibc. uClibc lacks some important things glibc has got (most notably the binary compatibility commitment), but is still perfectly *usable* (well, it supports most of POSIX by now, at least the parts people actually use).
It's more than GCC
Posted Apr 7, 2008 15:39 UTC (Mon) by coriordan (guest, #7544) [Link]
That, as Mr. Clinton will agree, depends on what the meaning of the word "is" is. I mean, sure, it *is* possible in a post to an LWN story. But where're the uClibc based distros? And I don't mean toy projects, unused experiments. /coriordan smokes a cigar
It's more than GCC
Posted Apr 7, 2008 18:05 UTC (Mon) by nix (subscriber, #2304) [Link]
Well, I know my systems wouldn't boot without it (it's damn useful as a libc in an initramfs: klibc is too small to handle a lot of stuff you might reasonably put in there, but uClibc happens just fine).
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 9:03 UTC (Fri) by bronson (subscriber, #4806) [Link]
> the total of the GNU project's code is much larger than the Linux kernel's size. Well, that might be relevant if anybody installed the total of the GNU project's code on their system. Has anybody compared something sensible, like the amount of code in a default Red Hat install? (insert standard "people say Linux simply because GNU/Linux looks awkward" argument here) I'm very curious... what companies are trying to push some sort of anti-GNU myth? Do you have any evidence to back this up? Why would any of them care enough to play mind games to somehow prevent their customers from hearing the GNU philosophy? That seems awfully unlikely doesn't it?
They call it GNU/Linux - thanks Sun!
Posted Apr 7, 2008 9:09 UTC (Mon) by coriordan (guest, #7544) [Link]
IBM is an example. They have TV ads saying that Linus wrote an operating system. It's not a big conspiracy constructing a delicate false world. It's a simple naming decision based on priorities. If they want to help the free software movement, they could call it "GNU/Linux" all or some of the time, and if they don't, they can just call it "Linux" and publish ads with fantasy tales of it's creation.
They call it GNU/Linux - thanks Sun!
Posted Apr 7, 2008 20:32 UTC (Mon) by bronson (subscriber, #4806) [Link]
Then this sounds like a false dichotomy...? Just because IBM doesn't particularly care about the GNU project, that doesn't mean that they're conspiring with other companies to prevent people from hearing about it either. Lies are pushed through TV adverts every day... If an IBM advert is the best indicator of an anti-free-software conspiracy, well, I don't think you have much to worry about. :)
They call it GNU/Linux - thanks Sun!
Posted Apr 7, 2008 21:14 UTC (Mon) by coriordan (guest, #7544) [Link]
Ah, maybe you misread my post. I said there's *no* conspiracy. IBM doesn't need to conspired with anyone to undermine the GNU project. A sufficient number of companies will do it of their own free will. The point is that most companies that profit from using, distributing, and providing services related to GNU software never tell the public or their customers that it is GNU software. The GNU project developed the software as a way to spread the philosophy, so when people spread the software while minimizing the link to the philosophy, they're undermining the GNU project. The GNU project works to give users freedom. So people undermining the GNU project are harming the cause of the free software movement. Sun decided to start acknowledging the GNU project. Fair play to them.
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 12:49 UTC (Fri) by Zack (guest, #37335) [Link]
All technical arguments aside, I call it GNU/Linux because rms politely asks me to. "Over time, calling the system “GNU/Linux” spreads awareness of the ideals of freedom for which we developed the GNU system. It is also useful as a reminder for people in our community who know about these ideals, in a world where much of discussion of free software takes a totally practical approach. When we ask you to call the system “GNU/Linux”, we do so because awareness of GNU slowly but surely brings with it awareness of the free software ideals of freedom and community." I think it's a valid request, and to me it's not overly cumbersome. Granted, at first it feels a bit strange in a conversation where the other parties are using "linux", but then again, the system being "strange" and "unorthodox" is what got me to run it in the first place.
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 11:41 UTC (Fri) by oblio (guest, #33465) [Link]
"The policy of calling the whole system "GNU" does beg the question - what proportion of the average free software-based system is actually GNU software? And of that software, how much of it can be considered part of the core of the system and how much as tools? Does the compiler suite count as a development tool or as part of the system core? Does the fact that it is required to build the system affect the equation? Does Gnome consider itself to be primarily part of GNU (is Solaris GNU?), or an independent project? Which proportion of free software systems (desktops or otherwise) use Gnome? And while we are at it, what proportion of such a system is actually Linux? (As a side note, I like to talk about the systems as being "Debian", "Ubuntu", "Gentoo" or whatever in order to sidestep the issue. That at least seems to me to be an uncontroversial description of them.)" ___ Furthermore, a lot of the most used F/OSS software is not GNU. MySQL is not GNU, PHP is not GNU, Apache is not GNU, Sendmail, Qmail, Postfix, Tomcat, KDE, OpenOffice, Inkscape, BIND, ZSH, PostgreSQL, SQLite,... are not GNU. GNU is a sort of glue these days, being the toolkit with which these applications are built. And except for GCC and GDB, everything is replaceable. Look at FreeBSD, NetBSD, OpenBSD, for example. It is an important part, and probably the most politically aware, but it is *not* the end-all-be-all some GNU supporters are portraying. A lot of people outside GNU are helping F/OSS too.
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 12:44 UTC (Fri) by grouch (subscriber, #27289) [Link]
Furthermore, a lot of the most used F/OSS software is not GNU. MySQL is not GNU, PHP is not GNU, Apache is not GNU, Sendmail, Qmail, Postfix, Tomcat, KDE, OpenOffice, Inkscape, BIND, ZSH, PostgreSQL, SQLite,... are not GNU.
Remove all of those and you can still have an operating system -- you can still interact with your computer. Remove either the GNU software or Linux and you don't have an operating system, in the typically-named "Linux" or "GNU/Linux" distribution. The kernel by itself is not an operating system. The shell, coreutils and essential libraries standing alone are not an operating system.
I run Debian GNU/Linux. I can strip out everything except GNU code and Linux code and still operate my computer. Everything I add to those increases the convenience and utility of my computer, but removing either of those leaves me with a powered paperweight. If my operating system did not already have a name, either "GNU/Linux" or "GNU+Linux" would be a logical choice as the two parts are essential to gain the characteristics of the operating system I prefer.
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 16:59 UTC (Fri) by oblio (guest, #33465) [Link]
"Remove all of those and you can still have an operating system -- you can still interact with your computer. Remove either the GNU software or Linux and you don't have an operating system, in the typically-named "Linux" or "GNU/Linux" distribution. The kernel by itself is not an operating system. The shell, coreutils and essential libraries standing alone are not an operating system." ___ And there are BSD licensed version for all of those, except GCC, GDB and a few others. My point is that you could probably run the Linux kernel with a BSD userland, and with software on top of that from outside the GNU ecosystem, and still have a full system. GCC is GNU's powerbroker :)
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 17:00 UTC (Fri) by oblio (guest, #33465) [Link]
Substitute "versions" with equivalent "applications/tools" :)
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 18:57 UTC (Fri) by JoeBuck (subscriber, #2330) [Link]
Currently, there isn't a BSD C library that works with the Linux kernel. There is uClibc, designed for embedded use (LGPL), as well as Cygnus newlib, also for embedded use (non-copylefted, BSD-like terms).But generally if you're on a desktop, laptop, or server, you rely on GNU libc.
They call it GNU/Linux - thanks Sun!
Posted Apr 4, 2008 20:08 UTC (Fri) by nix (subscriber, #2304) [Link]
Except, perhaps, when booting (some people use klibc in their initramfs/initrd; some use uClibc; nobody without disk space to throw away uses glibc).
They call it GNU/Linux - thanks Sun!
Posted Apr 5, 2008 3:27 UTC (Sat) by grouch (subscriber, #27289) [Link]
My point is that you could probably run the Linux kernel with a BSD userland, and with software on top of that from outside the GNU ecosystem, and still have a full system.
So, if I understand you correctly, your point is that there are probably other operating systems besides GNU/Linux. I can probably agree with that. ;)
I still prefer GNU/Linux, however, and have had no need for any other operating system for quite a few years. Any replacement would have to have at least as strong an emphasis on user freedoms. I've grown accustomed to being in control of my computer and data.
You are right - in a sense this IS the reason to use GNU/Linux
Posted Apr 5, 2008 14:44 UTC (Sat) by khim (subscriber, #9252) [Link]
And there are BSD licensed version for all of those, except GCC, GDB and a few others.
Yup. That's exactly why it's better to use GNU/Linux name! For example Android is Linux-based system but not GNU/Linux system. And since the GNU glue is gone it feels quite different then "normal" Linux (really GNU/Linux) system. The same it true for OpenWrt and other systems without GNU userland.
Once upon time the story way simple: we can omit GNU in GNU/Linux since it's there by default. It was even justifiable back then. But today people are talking about "triumph of the GNU on mobile phones" when they talk about Android exactly because they don't distinguish GNU and GNU/Linux. Today we have all four combinations in real use: nonGNU/nonLinux (like Windows), nonGNU/Linux (like Android), GNU/Linux (most Linuxes out there) and even GNU/nonLinux (like Debian GNU/kFreeBSD or Gentoo/FreeBSD). In many ways Debian GNU/kFreeBSD is closer to Debian GNU/Linux then to Android so GNU/Linux name makes perfect sense. And yes, the main reason to use GNU/Linux are two facts:
1. GNU and Linux make usable OS, anything less is not usable.
2. Either GNU or Linux can be replaced with something else.
They call it GNU/Linux - thanks Sun!
Posted Apr 15, 2008 23:05 UTC (Tue) by bignose (subscriber, #40) [Link]
> The policy of calling the whole system "GNU" does beg the question - what proportion of the average free software-based system is actually GNU software? That question isn't relevant to the request to call the operating system GNU. It deserves that name because that's what the creators of the operating system called it from the beginning, that's what its name has been for 25 years of effort toward making that operating system, and that's what they consistently and politely ask us all to call it. Linus is responsible for *one* program in a free software operating system: Linux. (Two, if you count 'git'.) Linux is a very important program, but it is *not* an operating system by itself. GNU, plus the program Linux, *is* an operating system, and it deserves to be called by the name its creators want us to call it. Calling it GNU/Linux acknowledges the fact that Linux is a very important part of the operating system, but it is still *one* program. No single program deserves the title of "operating system". The fact that it's an operating system at all, with thousands of parts crteated by many people but all managing to work together well, is because of the consistent efforts of the people creating GNU as an operating system.
Sun Microsystems' Next Linux Move (Seeking Alpha)
Posted Apr 3, 2008 23:54 UTC (Thu) by zooko (subscriber, #2589) [Link]
Hey, maybe Sun and Canonical will team up to maintain and develop Nexenta! That would be great. But I haven't heard any such rumors. :-(
I'd buy that for a dollar.
Posted Apr 4, 2008 0:45 UTC (Fri) by dmarti (subscriber, #11625) [Link]
What they really need is an OpenSolaris kernel package with "Linux kernel personality" that would install as a boot-time option on an existing Ubuntu system. Probably easier to add a fake Linux /sys and /proc at the kernel level than to clean up the crusty old Solaris userspace.
I'd buy that for a dollar.
Posted Apr 8, 2008 16:55 UTC (Tue) by captrb (subscriber, #2291) [Link]
They might be closer than you think. You can install certain linux distributions inside a Solaris zone. Unfortunately, the only supported one is kernel 2.4 only and therefore some really old distros. A grad student worked on a 2.6 version, but I don't believe it is stable. Sun may be working on it though.
How about SPARC version?
Posted Apr 4, 2008 1:50 UTC (Fri) by rogerd (guest, #4170) [Link]
Any chance at all of a Sun Linux that works on UltraSparc? My employer won't allow a PC on my desk. must use Sparc blade.
How about SPARC version?
Posted Apr 4, 2008 18:30 UTC (Fri) by larryr (guest, #4030) [Link]
Linux/SPARC can run inside a Logical Domain (LDom) running on Solaris/SPARC, if that helps; the native OS on the hardware would be a Sun supported OS at least, and there would be Linux running on the same computer, but not a Sun supported Linux I guess, for the moment at least.
Larry@Riedel.org