And here comes the Solaris 10 FUD from HP
Not even a day after the launch and already the FUD starts to flow. One recently brought to my attention was Martin Fink, at http://www.linuxcio.com/FBlogNov04.html. Martin Fink is HP’s Linux Vice-President.
Martin takes it on himself to inform folks about what Sun is not telling about Solaris 10/x86.
I’ll try to correct some of his misperceptions. The headings are below are Martin’s. I won’t quote the actual article, you can read it in his blog.
Is it really Solaris?
Martin’s argument here is that because SPARC is big-endian and x86 is little-endian, and that applications compiled for SPARC will not run on x86, that it can’t really be Solaris.
Sun has never claimed binary compatibility across different chip architectures (as against our claim of compatibility across multiple releases of the OS). I suspect Martin would be surprised to learn that Solaris on SPARC and Solaris on x86 are actually built from the same source tree… (you know, kind of like Linux for x86, linux for SPARC, linux for …).
He also makes a comment about moving data between Solaris SPARC and Solaris x86. I can only assume that he means physically moving disks as folks have been shipping information across networks for ages. Yes, if you try to move a disk with a ufs filesystem on it between the two architectures you will run into problems. However, ZFS is endian-neutral, so that problem goes away.
Is it really open?
Martin states that Sun has only ever talked about open sourcing Solaris on x86. Didn’t I already mention that Solaris is built from the same source tree? It’s going to be both.
While we have not decided on the actual license we will be using (which would explain why we have not announced it), there is commitment that the license will be OSI compliant.
Why I don’t think SUN will use the GPL
As I’ve already stated, OSI compliant. The GPL is not the be all and end all of open source licenses. It may be what we go for, it may not. We don’t know yet. As such I can’t really offer a lot of comment here, but he’s certainly entitled to his opinion.
Who can check-in code?
Unfortunately I’m involved in some of the discussion on this so there is not a lot that I can say, except that we would look pretty silly if we were to call it open and not allow external check-in to the open source code.
Linux inside; Well, maybe
Martin treats project Janus as an admission that there are more linux applicatoins out there than there are solaris applications. And this is news how?
It would be pretty arrogant not to acknowledge that there are some good linux applications out there that have not been built for Solaris. OK, so if we run these under Janus, we get the hardening that is Solaris as well as all of the cool stuff in 10 with regards to observability of tools like DTrace.
What we also get is server consolidation. Think about multiple instances of Janus inside Zones running on something as beefy as a V40Z; along with running your Solaris applications as well. Oh yes, you could also factor in the resource manager as well for sharing out the zones in a fair and predictable manner. Certainly sounds interesting to me.
Is Janus Indemnified?
Mu. That is, the question does not make sense.
We have not built a clone of a linux kernel, we have written to the kernel API. All of the code is our own. Indemnification does not enter into it.
Applications need developers
He continues to harp on about the admission of more applications available for linux here and tyhen goes on about how linux now has a collection of more than 9000 developr tools on freshmeat and another 12000 under development. I don’t know about you Martin, but thats a few too many for me. We and a lot of third parties have been developing for Solaris for a long time. We’re pretty happy with our toolset. You may not have noticed, but a lot of the tools that are made available for linux also actually compile and work under Solaris (there’s a shock, hey!).
He asks where the applications for Solaris are going to come from. Hmmm OK, off the top of my head how about we start with Oracle, Sybase, Veritas and Siebel. There’s a lot more, but I don’t have a list handy.
Where will the developer community come from?
Martin claims that there is no developer community for Solaris. Martin, you are sadly mistaken.
What problem is SUN trying to solve?
Martin makes some good points in here about what opening up code can achieve, however he is still under the delusion that Sun intends to control the codebase. Again he tries to say that Solaris SPARC and Solaris x86 are two entities. Didn’t I already point out that Solaris is built from the same source tree?
And who says we are trying to turn Solaris into Linux?
I would have thought that many of the benefits that he has already mentioned would be sufficient. We also know that there are folks out there who want to work with Solaris. The feedback we are seeing in the pilot programme certainly confirms this.
Ah, but what about TCO
In this topic, Martin questions the sustainability of Sun’s Solaris pricing plan. What he overlooks is that Sun is more than an Operating System company. Sun also sells hardware, support, middleware and applications. He again flogs the dead horse about Solaris SPARC and Solaris x86 being two different beasts. Do I have to say it again? They are built from the same source tree.
He also speaks about server shipments from various companies and how we compare, again questioning Sun’s viability. I won’t go into depth on the figures, it’s a huge paragraph. One thing I will mention though is that there already in excess of a million Solaris x86 licenses out there. Our demise has been predicted more than a few times in the past 23 years. Guess what? We’re still here.
He asks how many people are interested in running Solaris on x86 hardware. You only have to go back to when we deferred Solaris 9 to answer that one. The backlash was incredible.
In the end
I could say that if Martin doesnt understand our strategy, then that’s good for us and bad for HP. I think the main point that Martin has missed is that Sun is not an OS company. We are not a server company. We are not a storage company. We are not an applications or middleware company. We are a Solutions company. Unlike most of our competition we can provide a soltion from the hardware, through the kernel and middleware up into user space. We also don’t have to pay anyone in order to provide it, as we own the IP. OK, A customer doesn’t want to run on SPARC Solaris or x86 Solaris, why should we then not sell them hardware that is capable of running Linux or even (shock horror) Windows?
In a way I guess it’s predictable that the FUD has already started. I have faith that those who make the real decisions can tell the difference between FACT and FUD.
I mucked up the url for blastwave when I saved this. It’s right now.