OpenSolaris spork ready for download It is not quite ready for primetime, but with the announcement of OpenIndiana, a so-called spork of Oracle's OpenSolaris Unix distribution, the server world is getting a familiar, re-opened, and community-developed operating system aimed specifically at data center workloads. Alasdair Lumsden owns a hosting company in London …
...while some people are running scared and others are being smug, annoying platform evangelists, there's someone, somewhere, who is actually doing something about this mess. I like FreeBSD just fine and I'm damn glad it's there, but don't belittle this guy for having the chutzpah to flip the bird to Oracle and roll his own Solaris. If there weren't guys like him, you wouldn't have FreeBSD, either. Think about that.
I think it's awesome he's NOW putting some money and ingenuity and time into open source software. If El Reg was trying to highlight that this is in fact what open source is about - not just complaining when Oracle decides to take its bat and ball away and go home - then good on 'em.
But really - 50 servers running for paying customers, and he hasn't already been factoring in $1000 per socket? Open Source Software (Free Software) is not about getting away without paying - it's about having the freedom to improve and customise the software. The guy says he was contributing previously - I guess at some significantly lower level (or else the article is hardly news is it?).
Over 6 years I've invested roughly $750K gross (errm $500K net?) into the F&OSS I run my business on. To me it seems that just because licensees don't have a legal obligation to pay/contribute they think they can get away without it - it's all "someone else will pay". I'm sorry to burst someone's utopian dream - if nobody pays then the programmers eventually get tired of being evicted from their houses, and no young programmers have success stories to inspire them to pursue this model.
Perhaps if this guy had been contributing at this level all along then Sun would still be around, or the new sponsor (Oracle) would have felt they were getting real value out of having their asset free/open - not just simply carrying a lot of freeloaders with no net benefit.
It looks like his business is in hosting services for customers, this is a very cut throat business with extremely thin margins... Free (as in cost) software is pretty much essential in this business, if you pay for your software you won't be price competitive against your competitors who don't.
If everyone else was building houses on sand would you too? Or do you build on rock? Sure building on rock is more expensive - but after the first rains come yours is the only one left standing. I had the choice 7 years ago whether to build by business on proprietary software, on F&OSS with investment or on F&OSS with none or minimal investment. I've spent as much (probably more) investing in the F&OSS that my business runs on than on comparable proprietary software. Sure there has been little rain so it may seem that the only difference between my business and others is that I've made less profit - but rain always eventually comes.
IANAL but I did work at Sun on both OpenSolaris and OpenJDK launches.
If you accept the code to OpenSolaris under the CDDL, my understanding is that you are covered by a very explicit patent grant in that license. The GPLv2 under which OpenJDK was released has similar (weaker?) provisions, and implementations that conform to Java specs also have some degree of patent coverage. However, Google didn't take the OpenJDK code, and the product that they released does not conform to a Java spec. The Google/Java situation is very different to a fork of OpenSolaris code.
Mine's the one with the elbow patches.
How much would that have cost? If those are the major concerns, then it might've been smart to see just how much it would've cost to get those features into an OS that already has lots of contributors and has been shipping production releases for years. It might have been prohibitively expensive, but there's no mention of that option at all so how can we be sure it was even considered?
The OpenSolaris licence (CDDL) isn't compatible with the GPL -- there's a "deadly embrace" situation, with neither licence permitting additional restrictions to be imposed on distribution.
This means that a Linux kernel compiled with ZFS support would not be legally redistributable; as you would be either out of compliance with the OpenSolaris licence if you distributed it under GPL, or out of compliance with the GPL if you distributed it under CDDL.
The only way you could package a ZFS kernel as a .deb or .rpm would be to have the package depend on the build toolchain, install the source code somewhere sane, and do the actual configure and build operations in the post-install script. That way, the resulting binary isn't being distributed, so neither licence is invoked.
FreeBSD is a very nice OS but in my humble opinion it is not production ready unless you really feel like spending lots of time tinkering.
FreeBSD is suffering from a huge lack of "eyeballs" compared to Linux. Over the past years FreeBSD have missed many release schedules and you just get the impression that there are not enough FreeBSD developers out there to deliver the functionality and stability which we see in Linux distributions such as Redhat.
An important production issue is lack of official hardware support. Most vendors of servers, disk controllers, etc, will officially support Linux but not FreeBSD. The same goes for software. It is sometimes diffult to find FreeBSD support for your favourite configuration management tool, virtual machine hypervisor, etc. etc.
Many databases, such as MySQL, work better on Linux because the majority of the developers, consultants, etc. work on Linux. So the Linux platform gets tested, documented and optimised much better.
Package management on FreeBSD is less mature than Linux and due to lack of official FreeBSD support you often end up compiling your own packages (and kernel if you have a bad day) and maintaining your own package repositories. This really slows down patching/vulnerability management.
So in many FreeBSD production environments you may see low productivity and delayed projects because your staff will be spending time on debugging and fixing some low level critical issue that just works out of the box on Linux. Well, that is my experience. Your mileage may wary.
May your OS be with you!
..just shell out the money Mr Larry demands from Solaris users now (or in the future).
If you use BSD, you should use Postgres, not that MySQL "random row loser". And you should also not expect that commercial bloat like VMWare being ready with all those strange commercial features.
Instead, use the strengths of the BSD ecosystem and adapt to it. Otherwise you will be disappointed with 100% probability. Ever head of BSD Jails, btw ?
Use Jails, Python, Postgres, PHP, Perl, OpenLDAP on BSD. If you want to run Mono, ActiveDirectory. MySQL in a VMWare instance on BSD you might get quite bad results. But that's your fault.
And please don't bank on a promised feature. Only plan with features which are already there and considered stable. Forget everything else. Forget that notion that "there must be a frequent improvements release cycle" or other corporate blurbs from the windows/VMWare/Oracle world.
If you want dancing monkeys, go with Mr Bill or Mr Larry. If you want stability, use BSD or Linux.
The problem isn't that Oracle are charging $1000/socket now - it is what it will do a few years down the line that should be the real concern. Sun never excelled at vendor lock-in (some may call that a virtue); Oracle on the other hand exemplify it. The per-socket prices now are just a taster of what is to come once they have their customers more firmly welded to their OS.
Talk to your DBAs about Oracle pricing plans - you may want to bring a box of tissues and your sympathetic face. And don't mention yachts.
This post has been deleted by its author
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
