You're missing some stuff in the list that I think should fit under what a Joe User needs.
The big thing is 3rd party vendor support, for even ubiquitous things like broadband, or devices like printers, video cards, wifi, memory sticks, cameras, etc. Many of these are not officially supported, have varying degrees of usability under linux (usually pretty terrible).
As much as I like free software, it doesn't jive with propietary formats like DVD css, mp3, and others. Makes it a hassle, or even legally shady to get these working.
Even if true, it is a stupid allegation. Intel don't have a monopoly on x86 compilers, there are probably more x86 apps built with gcc and MS's compilers.
If AMD don't like Intel's compilers they are free to make their own compiler or work harder on improving gcc, pathscale or whatever instead of whining about it.
These are probably RDMA capable which should save CPU because of some offload capability plus zero-copy unlike TCP/IP which needs to transfer the buffer between the user memory into kernel buffers.
The type of customers for this would be database clusters and scientific clusters mostly.
My guess is for applications that use only TCP/IP there won't be much if any performance gain, but if the app is RDMA-aware it could be boosted by this.
But you likely had to install Mepis, and perhaps
get dvdcss separately (most distros don't ship this for legal reasons), configure dma, etc. It's not like you bought the toshiba with all this pre-configured and supported..
If I buy a laptop, I expect it to play my DVDs out of the box regardless if it is Linux or Windows pre-installed on it. Windows has the advantage here because there are multiple legal players available for it, and NONE (afaik) for Linux.
I'm saying that if HP plan on selling this to home users I hope they have some problems like this solved.
I have DMA enabled. I didn't have this problem with kernel 2.4 (at least to this degree). The base problem I started hitting was libdvdread couldn't read past some certain block in cases and I think it was related to how my IDE device was being accessed by kernel. libdvdread hasn't been updated in ages!
I was able to get around it for a short while by emulating SCSI on my IDE drive but a couple of kernel security updates later through YAST and SCSI emulation breaks totally. XINE hangs, vlc crashes, etc.
Also, depending on what DVD I watch my luck varies (it can play some, play portions of others, not play others at all) and I don't own any unencrypted dvds either.
For example I have never been able to play my Attack of the Clones DVD in a satisfactory manner on Linux (even with 2.4).
A laptop that doesn't play encrypted DVD's would be a showstopper for me.
And even getting legally shady alternatives like decss to work with the 2.6 kernel on my desktop has been an adventure in futility (at least for most of my dvds, where they don't play at all, or don't play past certain chapters and so on).
Are there any F/OSS apps we non-Mac users can use to play with this?
Likely possible through some indirect manner and some additional work. I don't know garageband, but I suspect it can export individual tracks to mp3, etc. Get someone to do this, and re-assemble the tracks in an Audacity project.
This is really interesting in terms of possibilities. Imagine having some Qt type license for music where if you released a song under this license, another artist could modify it freely and redistribute it provided they also have it in the same multi-track format for future revision. Open music. Collaboration by anyone.
Thank-you. An insightful post in a sea of regular slashdot stupidity.
Red Hat and Suse (and I mean the non-cheap Enterprise Editions) are your only option if you want to run anything halfway useful: BEA, Oracle, DB2, Websphere, SAP, etc.
Your database server corrupted your data? Oh, you running it on Slackware? Piss off! Call us back when you reproduce the problem on SLES 9.
Seriously, like 95% of the people on here must be running "servers" out of their basement or something...
Sun makes it sound like a simple recompile is all that is needed. Of course this is NOT the case if your software is sufficiently advanced like DB2. Secondly, a port requires a new set of tools which means a new set of unknown problems just waiting to be discovered (from the OS, to the compiler, to libc which if it were really worth much in terms of performance would have been hand-coded in assembly anyway and prone to bugs when going from sparc to x86!)
That's just porting... nevermind testing, packaging, documenting, supporting, maintaining. This costs $$ in terms of hardware and employees.
If Sun were serious about their "anti-competitive" allegations towards IBM, pony up the hardware and $$$ IBM needs to do the ports so that IBM can at least re-coup its development costs. What's it going to cost $5M, $10M? If Sun were serious, and lack of IBM software were truly an inhibitor to their sales, they should consider this an investment.
Moving optimizations into software is the wave of the future for 2 reasons:
1. On-chip circuitry to handle things like out of order, dynamic register renaming, branch prediction, and other fancy items seen in contemporary processors add to power consumption and thermal requirements. The next real trend in CPUs is the multi-core which means that the individual cores better be lean!
2. It is far easier to propagate optimizations in software than it is in hardware. Want to add a new architectural feature (eg, change the set size of architected registers)? You can do it in the compiler logic at much less cost than having to fab it in hardware.
The multicore sparcs have already shortened their pipeline and removed hardware branch prediction. Power is doing similar stuff.
My understanding of itanium is that it was designed to be the successor to PA-RISC 2 (which was already 64bit). Yes, instructions aren't 32bit and you require a 64bit kernel to boot off of itanium, but the limitations to using 32bit software ends THERE. In fact, AFAIK, you require a 64bit kernel of the OSes on the other chips SPARC, AMD64, PPC, to have a hybrid 32/64 bit computing environment. Otherwise, how come I can't compile a 64bit windows app and run it on AMD64 with regular 32bit Win XP? I guess AMD aren't so clever afterall, and only doing what every other chip designer has ever done.
HP-UX provides an ILP32 environment (meaning 32bit using ia64 ISA instructions) on itanium so that porting from PA-RISC is easy if your app is still not 64bit clean. And 32bit itanium binaries are first class citizens, running completely optimally and being used for CPU benchmarks.
Itanium is not a bad chip by any means. The only thing that is going to hurt it is that it is trying to be everything to everyone.. It has to keep the PA-RISC folks happy, the x86 folks happy, the Alpha folks happy, the Tandem Non-Stop folks happy (next itanium will even have lockstep for example) and just the scope of what it's supposed to be -- nothing short of revolutionary, is why everyone thinks its a disappointment. No other processor has had this kind of pressure and been put under this large a microscope.
Do you mean the itanium sucks because it can't do another instruction set as fast as its own? How fast does Alpha, Mips, PPC, and Sun run x86 code?
Itanium runs 32bit software plenty fast. HP's SPEC scores are all with 32bit compiled binaries on HP-UX which fully supports an ILP32 environment. Smaller data structures == better cache utilization. Oh yeah... Linux sucks and doesn't have this support.. oh well...
And this is what I don't get about the continuous bashing of itanium especially from an open source savvy crowd. From a Linux perspective, what the hell is the issue? Can't you just recompile your open source code on itanium? Isn't this what people do anyway on Gentoo??
Out of any applications, I would say games are inherently the least portable than anything else. This is why you see a ton of games on Windows 32bit but nowhere else.So I doubt a simple recompile is even feasible first of all. Also since you can't mix and match object code of different word size on any platform (even Opteron) you would require the entire runtime environment and associated DLL's to be available in 64bit mode like DirectX, etc.
But supposing we do have a 64bit game, will it run faster? That depends:
1. Did the on-disk data structures expand? If so, how good is your I/O subsystem (Opteron probably has the disadvantage here at least over Sparc boxes)
2. Did in-network data structures expand? If so, how good is your network subsystem (see point 1 about Opteron vs Sparc)
3. Was data alignment perserved when the structures expanded?
4. Did it take advantage of increased virtual memory and shared memory? Do you have enough RAM on your machine so that you don't start paging?
5. Does it have issues with data-cache misses because of larger data types? How big a cache you got?
Registers aren't the only thing in the equation...
Oh, I am living under a rock... I will bite troll.
Opteron is already mainstream.
Of course it is because Overclockers UK says so. Nevermind that Windows isn't even released for it! Oh yes, Linux x86-64 is mainstream, I forgot.
10000 Opterons in an SMP?? You have any clue what you are talking about??? Show me a link for this (I'm not talking clusters either because they can be made with anything) instead of spewing AMD fanboy crap.
I remember the same thing happened when Athlon was first released and was first to 1GHz and would beat pentium in all the benchmarks for a cheaper price. I also remember Intel coming back strong and eventually winning the x86 performance race.
Your prediction is wrong. It will take much longer than next year (if ever) when Opteron starts to become "mainstream". The biggest reason is application availablility along with essentially non-existant tier-1 vendor support.
I can list about 4 software products that I know of that currently take advantage of the AMD64 instruction set (two of them are OS's, the main one not even released yet!). Currently, there is no AMD64 java, no *good* AMD64 compilers -- MS or Sun will likely have these by end of next year, but may take even longer to mature AND the Sun compiler will be Solaris only.
No corporation is going to spend their budget on new AMDs under these circumstances. The development environment well.. sucks and what's the point of upgrading your hardware if you only want to run 32bit x86 binaries?
As for scaling to multi-way servers, etc. Show me an example? If a corporation is going to spend on large SMP boxes, I doubt x86 compatability is at the top of their list of priorities. They will already have Power, Sparc, or IA64 big iron.. Yes, Itanium. Which is already in 64-way boxes and soon to be in 128-way boxes. Show me such scalability with an opteron!
Buy a clue (maybe get this cheaper in China) or better yet RTFA.Particularly:
What's more, many Chinese DVD manufacturers don't pay the $10 to $15 in royalties due per unit for patented technologies -- penalizing established consumer-electronics companies that honor intellectual-property rights
How the HELL is a company that operates from a country without a joke legal system supposed to compete? This is simply *stealing* technology and selling it cheap because you didn't have to invest in the original R&D. Are you saying since we can't compete with their stealing, we the west should concede defeat or try to compete by removing our patent laws? Who will invest in innovation then? Research doesn't come cheap.
And of course we can get into unhealthy slave like work conditions blah blah, etc.
Care to back up the integer performance claims with a link?
I'm looking at spec.org and the SPECINT latest numbers between the closest comparable Opteron and Itanium2 systems are the 4 way results. 4 x 2000MHz Opteron get lower integer performance rates than 4 x 1500 MHz Itanics.
Opteron 4way Itanium2 4way
Also explain to me why you think a webserver is
(a) necessarily CPU bound
(b) Is performant dependent on integer cpu performance.
I would think a typical webserver is I/O bound. In the case of a "cpu bound" webserver likely your bottleneck is OS scheduling or other synch primitives in the OS. Disk subsystem and Memory will also dictate webserver performance.
Shouldn't that read "new TPC-H Record Set Using Itanium2?"
These "benchmarks" are hogwash, they don't measure one vendor's database against the others in general. They measure overall solution where hardware plays the key role. And they make good proof of concepts. Not to mention other politics (and discounts) in the mix... Vendors need to pick and choose their benchmarks so as to not outdo their other strategic benchmarks.
The only interesting benchmark on that whole website in respect to the db vendor are the Oracle and DB2 pSeries AIX results on TPC-C because they run on basically identical hardware. For almost 4% increase in Price/tpmC Oracle gets you less than 1% performance improvement. From which I can conclude that Oracle and DB2 are basically the same on similar hardware with DB2 possible edging out on cost/performance.
I also have yet to see a good Linux SMP result on a large box (or even clustered for that matter). Unix and Windows have them. When linux can play with these guys in the high-end TPC-C and TPC-H then this will be interesting.
I don't know about "cheaper". If the account were worth it, Microsoft could afford to undercut RH's prices for long term revenue. I doubt Redhat would do the same...
At my company, we develop software that runs on multiple platforms. We are considered "partners" by the major Unix and Linux providers yet Redhat is the ONLY one of these companies that insists on us paying for their distro/solution in order to do software development that adds value to their product. Go figure.. If I were the PHB, I would have dropped support for them.
My point is Redhat is entering a very dangerous market with an arrogant attitude.
Once upon a time I thought the whole selling point
of linux was you could buy a distro package from a vendor and deploy on all your servers and have great savings in $$$, etc.
Now to run any enterprise app, you need an enterprise distro that is a per cpu or system cost so what is the selling point? Also doing things like modifying the kernel source is not supported so what good is "free as in speech"? You might as well run windows on your cheap hardware.
As for expensive hardware, what? Does Red Hat think they can compete with IBM, HP, Sun in industrial strength UNIX operating systems?
Forget about the competition with Suse (who in MHO has a superior product), this spells the beginning of the end of a once great company I'm afraid.
"32 bit" is more a function of the OS and compilers
than it is of the processor. For example, I use HP-UX on itanium which is a hybrid 32/64 OS and whose compilers generate both modes natively in the instruction set. Don't blame shortcomings in popular operating systems and legacy code on the Itanium.
Our company ports our software to Sun, AIX, HP-UX, and Linux among other platforms. Some of my gripes about the Linux port (and the GNU tools in general i.e. gcc, gdb, etc):
1. If we hit a problem, the likelyhood of it being fixed any time soon is just not there (contrary to what some people think).
Eg. We found a scenario where attempting to attach to a shared memory segment on Linux caused a segfault. We notified the gnu developers about this and they agreed that it was a problem. However they wanted to wait on the fix because it was a bit risky to put in! We even helped them by providing a patch to fix it. I mean it frickin traps!!! If this happened on HP or Sun we would get a fix in a week tops.
2. Requesting features. No, we don't have the R&D resources to be spending time on writing an IA64 stacktracer. Commercial Unixes have this. The Gnu community says stuff like we'll get around to it eventually. Or asking the gcc folk to have a particular compile option on ia64 actually work -- their response -- either do it yourselves or pay us to do it. ??? Right.. We now pay Intel for their working compiler.
3. Keeping a support matrix of what distros your software works on, etc. This is a hassle as there are different kernel levels, libc levels, abi's, and what have you. There's even more variance as some distros provide additional kernel hacks that aren't in the official linux kernel. This kind of thing keeps you on your toes.
Don't get me wrong, Linux is great, I use it on my desktop and love it for its flexibility but there are some situations where having a large company backed OS that comes at a cost is actually worth it. The right tool for the job I guess..
What really happened (from one of Kurt's biographies) is that Albini was highly sought after by Kurt for work on In Utero because of some of his earlier recordings -- Pixies' Surfer Rosa for example...
Kurt wanted and liked the raw signature sound of Albini's -- it was the record company that hated it. And there was a lot of pressure to have it re-recorded and remixed, but both Albini and Nirvana resisted. Essentially only Heart-Shaped Box and All Apologies were remixed by Andy Wallace (same guy as on Nevermind).
spinner.com has a pretty amazing variety of music. I would go there first.
Also, participate on messageboards for your favourite artists, other fans will give you ideas of what's cool and new, etc.
Slashdot Mirror
The big thing is 3rd party vendor support, for even ubiquitous things like broadband, or devices like printers, video cards, wifi, memory sticks, cameras, etc. Many of these are not officially supported, have varying degrees of usability under linux (usually pretty terrible).
As much as I like free software, it doesn't jive with propietary formats like DVD css, mp3, and others. Makes it a hassle, or even legally shady to get these working.
If AMD don't like Intel's compilers they are free to make their own compiler or work harder on improving gcc, pathscale or whatever instead of whining about it.
The type of customers for this would be database clusters and scientific clusters mostly.
My guess is for applications that use only TCP/IP there won't be much if any performance gain, but if the app is RDMA-aware it could be boosted by this.
But you likely had to install Mepis, and perhaps get dvdcss separately (most distros don't ship this for legal reasons), configure dma, etc. It's not like you bought the toshiba with all this pre-configured and supported..
If I buy a laptop, I expect it to play my DVDs out of the box regardless if it is Linux or Windows pre-installed on it. Windows has the advantage here because there are multiple legal players available for it, and NONE (afaik) for Linux.
I'm saying that if HP plan on selling this to home users I hope they have some problems like this solved.
I have DMA enabled. I didn't have this problem with kernel 2.4 (at least to this degree). The base problem I started hitting was libdvdread couldn't read past some certain block in cases and I think it was related to how my IDE device was being accessed by kernel. libdvdread hasn't been updated in ages!
I was able to get around it for a short while by emulating SCSI on my IDE drive but a couple of kernel security updates later through YAST and SCSI emulation breaks totally. XINE hangs, vlc crashes, etc.
Also, depending on what DVD I watch my luck varies (it can play some, play portions of others, not play others at all) and I don't own any unencrypted dvds either.
For example I have never been able to play my Attack of the Clones DVD in a satisfactory manner on Linux (even with 2.4).
A laptop that doesn't play encrypted DVD's would be a showstopper for me.
And even getting legally shady alternatives like decss to work with the 2.6 kernel on my desktop has been an adventure in futility (at least for most of my dvds, where they don't play at all, or don't play past certain chapters and so on).
Likely possible through some indirect manner and some additional work. I don't know garageband, but I suspect it can export individual tracks to mp3, etc. Get someone to do this, and re-assemble the tracks in an Audacity project.
This is really interesting in terms of possibilities. Imagine having some Qt type license for music where if you released a song under this license, another artist could modify it freely and redistribute it provided they also have it in the same multi-track format for future revision. Open music. Collaboration by anyone.
Red Hat and Suse (and I mean the non-cheap Enterprise Editions) are your only option if you want to run anything halfway useful: BEA, Oracle, DB2, Websphere, SAP, etc.
Your database server corrupted your data? Oh, you running it on Slackware? Piss off! Call us back when you reproduce the problem on SLES 9.
Seriously, like 95% of the people on here must be running "servers" out of their basement or something...
Sun makes it sound like a simple recompile is all that is needed. Of course this is NOT the case if your software is sufficiently advanced like DB2. Secondly, a port requires a new set of tools which means a new set of unknown problems just waiting to be discovered (from the OS, to the compiler, to libc which if it were really worth much in terms of performance would have been hand-coded in assembly anyway and prone to bugs when going from sparc to x86!)
That's just porting... nevermind testing, packaging, documenting, supporting, maintaining. This costs $$ in terms of hardware and employees.
If Sun were serious about their "anti-competitive" allegations towards IBM, pony up the hardware and $$$ IBM needs to do the ports so that IBM can at least re-coup its development costs. What's it going to cost $5M, $10M? If Sun were serious, and lack of IBM software were truly an inhibitor to their sales, they should consider this an investment.
1. On-chip circuitry to handle things like out of order, dynamic register renaming, branch prediction, and other fancy items seen in contemporary processors add to power consumption and thermal requirements. The next real trend in CPUs is the multi-core which means that the individual cores better be lean!
2. It is far easier to propagate optimizations in software than it is in hardware. Want to add a new architectural feature (eg, change the set size of architected registers)? You can do it in the compiler logic at much less cost than having to fab it in hardware.
The multicore sparcs have already shortened their pipeline and removed hardware branch prediction. Power is doing similar stuff.
HP-UX provides an ILP32 environment (meaning 32bit using ia64 ISA instructions) on itanium so that porting from PA-RISC is easy if your app is still not 64bit clean. And 32bit itanium binaries are first class citizens, running completely optimally and being used for CPU benchmarks.
Itanium is not a bad chip by any means. The only thing that is going to hurt it is that it is trying to be everything to everyone.. It has to keep the PA-RISC folks happy, the x86 folks happy, the Alpha folks happy, the Tandem Non-Stop folks happy (next itanium will even have lockstep for example) and just the scope of what it's supposed to be -- nothing short of revolutionary, is why everyone thinks its a disappointment. No other processor has had this kind of pressure and been put under this large a microscope.
Do you mean the itanium sucks because it can't do another instruction set as fast as its own? How fast does Alpha, Mips, PPC, and Sun run x86 code?
Itanium runs 32bit software plenty fast. HP's SPEC scores are all with 32bit compiled binaries on HP-UX which fully supports an ILP32 environment. Smaller data structures == better cache utilization. Oh yeah... Linux sucks and doesn't have this support.. oh well...
And this is what I don't get about the continuous bashing of itanium especially from an open source savvy crowd. From a Linux perspective, what the hell is the issue? Can't you just recompile your open source code on itanium? Isn't this what people do anyway on Gentoo??
Out of any applications, I would say games are inherently the least portable than anything else. This is why you see a ton of games on Windows 32bit but nowhere else.So I doubt a simple recompile is even feasible first of all. Also since you can't mix and match object code of different word size on any platform (even Opteron) you would require the entire runtime environment and associated DLL's to be available in 64bit mode like DirectX, etc.
But supposing we do have a 64bit game, will it run faster? That depends:
1. Did the on-disk data structures expand? If so, how good is your I/O subsystem (Opteron probably has the disadvantage here at least over Sparc boxes)
2. Did in-network data structures expand? If so, how good is your network subsystem (see point 1 about Opteron vs Sparc)
3. Was data alignment perserved when the structures expanded?
4. Did it take advantage of increased virtual memory and shared memory? Do you have enough RAM on your machine so that you don't start paging?
5. Does it have issues with data-cache misses because of larger data types? How big a cache you got?
Registers aren't the only thing in the equation...
Opteron is already mainstream.
Of course it is because Overclockers UK says so. Nevermind that Windows isn't even released for it! Oh yes, Linux x86-64 is mainstream, I forgot.
10000 Opterons in an SMP?? You have any clue what you are talking about??? Show me a link for this (I'm not talking clusters either because they can be made with anything) instead of spewing AMD fanboy crap.
I remember the same thing happened when Athlon was first released and was first to 1GHz and would beat pentium in all the benchmarks for a cheaper price. I also remember Intel coming back strong and eventually winning the x86 performance race.
I can list about 4 software products that I know of that currently take advantage of the AMD64 instruction set (two of them are OS's, the main one not even released yet!). Currently, there is no AMD64 java, no *good* AMD64 compilers -- MS or Sun will likely have these by end of next year, but may take even longer to mature AND the Sun compiler will be Solaris only.
No corporation is going to spend their budget on new AMDs under these circumstances. The development environment well.. sucks and what's the point of upgrading your hardware if you only want to run 32bit x86 binaries?
As for scaling to multi-way servers, etc. Show me an example? If a corporation is going to spend on large SMP boxes, I doubt x86 compatability is at the top of their list of priorities. They will already have Power, Sparc, or IA64 big iron.. Yes, Itanium. Which is already in 64-way boxes and soon to be in 128-way boxes. Show me such scalability with an opteron!
What's more, many Chinese DVD manufacturers don't pay the $10 to $15 in royalties due per unit for patented technologies -- penalizing established consumer-electronics companies that honor intellectual-property rights
How the HELL is a company that operates from a country without a joke legal system supposed to compete? This is simply *stealing* technology and selling it cheap because you didn't have to invest in the original R&D. Are you saying since we can't compete with their stealing, we the west should concede defeat or try to compete by removing our patent laws? Who will invest in innovation then? Research doesn't come cheap.
And of course we can get into unhealthy slave like work conditions blah blah, etc.
I'm looking at spec.org and the SPECINT latest numbers between the closest comparable Opteron and Itanium2 systems are the 4 way results. 4 x 2000MHz Opteron get lower integer performance rates than 4 x 1500 MHz Itanics. Opteron 4way
Itanium2 4way
Also explain to me why you think a webserver is
(a) necessarily CPU bound
(b) Is performant dependent on integer cpu performance.
I would think a typical webserver is I/O bound. In the case of a "cpu bound" webserver likely your bottleneck is OS scheduling or other synch primitives in the OS. Disk subsystem and Memory will also dictate webserver performance.
Shouldn't that read "new TPC-H Record Set Using Itanium2?" These "benchmarks" are hogwash, they don't measure one vendor's database against the others in general. They measure overall solution where hardware plays the key role. And they make good proof of concepts. Not to mention other politics (and discounts) in the mix... Vendors need to pick and choose their benchmarks so as to not outdo their other strategic benchmarks. The only interesting benchmark on that whole website in respect to the db vendor are the Oracle and DB2 pSeries AIX results on TPC-C because they run on basically identical hardware. For almost 4% increase in Price/tpmC Oracle gets you less than 1% performance improvement. From which I can conclude that Oracle and DB2 are basically the same on similar hardware with DB2 possible edging out on cost/performance. I also have yet to see a good Linux SMP result on a large box (or even clustered for that matter). Unix and Windows have them. When linux can play with these guys in the high-end TPC-C and TPC-H then this will be interesting.
At my company, we develop software that runs on multiple platforms. We are considered "partners" by the major Unix and Linux providers yet Redhat is the ONLY one of these companies that insists on us paying for their distro/solution in order to do software development that adds value to their product. Go figure.. If I were the PHB, I would have dropped support for them.
My point is Redhat is entering a very dangerous market with an arrogant attitude.
Now to run any enterprise app, you need an enterprise distro that is a per cpu or system cost so what is the selling point? Also doing things like modifying the kernel source is not supported so what good is "free as in speech"? You might as well run windows on your cheap hardware.
As for expensive hardware, what? Does Red Hat think they can compete with IBM, HP, Sun in industrial strength UNIX operating systems?
Forget about the competition with Suse (who in MHO has a superior product), this spells the beginning of the end of a once great company I'm afraid.
This is apples and oranges IMO.
Good point! Look what happened when a dying and defeated Informix tried to take on IBM in a legal patent battle... They got bought out a year later.
1. If we hit a problem, the likelyhood of it being fixed any time soon is just not there (contrary to what some people think).
Eg. We found a scenario where attempting to attach to a shared memory segment on Linux caused a segfault. We notified the gnu developers about this and they agreed that it was a problem. However they wanted to wait on the fix because it was a bit risky to put in! We even helped them by providing a patch to fix it. I mean it frickin traps!!! If this happened on HP or Sun we would get a fix in a week tops.
2. Requesting features. No, we don't have the R&D resources to be spending time on writing an IA64 stacktracer. Commercial Unixes have this. The Gnu community says stuff like we'll get around to it eventually. Or asking the gcc folk to have a particular compile option on ia64 actually work -- their response -- either do it yourselves or pay us to do it. ??? Right.. We now pay Intel for their working compiler.
3. Keeping a support matrix of what distros your software works on, etc. This is a hassle as there are different kernel levels, libc levels, abi's, and what have you. There's even more variance as some distros provide additional kernel hacks that aren't in the official linux kernel. This kind of thing keeps you on your toes.
Don't get me wrong, Linux is great, I use it on my desktop and love it for its flexibility but there are some situations where having a large company backed OS that comes at a cost is actually worth it. The right tool for the job I guess..
What really happened (from one of Kurt's biographies) is that Albini was highly sought after by Kurt for work on In Utero because of some of his earlier recordings -- Pixies' Surfer Rosa for example...
Kurt wanted and liked the raw signature sound of Albini's -- it was the record company that hated it. And there was a lot of pressure to have it re-recorded and remixed, but both Albini and Nirvana resisted. Essentially only Heart-Shaped Box and All Apologies were remixed by Andy Wallace (same guy as on Nevermind).
spinner.com has a pretty amazing variety of music. I would go there first. Also, participate on messageboards for your favourite artists, other fans will give you ideas of what's cool and new, etc.