Do you have questions about the site? Write to bruce at perens.com .
List of all articles we've ever published.
Source Code
Developer's Documentation
RSS
XML Sitemap
XML News Sitemap
![[Valid RSS]](/images/valid-rss.png "Validate my RSS feed")
Sun's Latest Follow-The-Leader Project
Sun has a project, "Indiana", which is going to make Solaris more GNU/Linux-like, not just in user experience but also for deeper levels such as system programming and device drivers. The article that Stephen Shankland wrote at ZDNet brings up things that some might find humorous, besides the interesting project itself.
Back about seventeen years ago, I was deeply into Computer Aided Design and Computer Aided Engineering, and was just getting into all the fun global networks like BITNET, the ARPANET/Internet, TYMNET, etc. that were humming away at the time (with various gateways between them). The engineering group I was a member of was going to Sun Sparcstations with a BSD Unix called SunOS on them, instead of PCs running DOS and a couple trying out Windows 3.
There were many ways the Sun Way was so much cooler than what was going on with our DOS and Windows 3 based i386 and i486 boxes as we moved the more willing users (strangely also the more productive and quick to learn ones) from 1990 to 1992. Those PCs needed all kinds of optional hardware accessories and software utilities to make them useful, and took a lot of two guy's time to administer. Contrasted sharply with Sparcstations, right out of the box, purchased with a single part number, the Suns had cool hardware like high res graphics, networking, mouse. Devices didn't require special jumpers and hairy configuration like the PCs, just plug in the stuff and, zoom, up and running. A boot PROM with some nonvolatile RAM that held FORTH firmware and software, modifiable, to diagnose problems, change boot devices and other device parameters and how the OS would see devices on the fly; heck could even freeze the OS and drop down into PROM mode and then start OS again. The software to run and make the machine go was even better, the SunOS had built in stuff for remote administration (NIS, rsh, rcp, rtar), backup (tar and dump), file sharing (NFS), email (sendmail), graphics and gui (X11R4, OpenLook with Motif) which was exciting to me as part of my job was to procure and install and configure software. Another part of my job was to customize and create needed integration pieces for all the other in house and client systems we had to work with, and SunOS included a compiler and libraries and the API to the operating system was open and documented. Not to mention our engineering applications ran twice as fast on the bad boy Sun gear as the fastest of the wimpy clunky PCs.
So I was a Sun fanboy for computing "at work", Sun was blazing the path for what a real "workstation" should be in hardware, firmware, OS and utility software from a one-stop shop. Totally butt-kicking for the low and midsized tiers of tech computation compared to what IBM, HP, Compaq, and DEC had going on at the time (we'll talk about NeXT, Apple, SGI at some other time, different markets for those)
Sun had imitators to flatter them, there was the DEC Vaxstation and DECStation, HP had the Apollo / 3000 series, IBM and Sony and others also made workstations. Sometimes each of these vendors would make a machine that work outperform a Sun box, but Sun would always jump away from them, and running benchmarks on real world apps that used the whole machine, i.e. more than just CPU and memory cache with MIPS, FLOPs and such, Sun would invariably take the prize.
Sun grew their product line from performance technical networked desktop towards the high end of serverdom until they were in 1998 the number one Unix server, THE way (outside of mainframe scale) to run Oracle and business applications and services. Sun made standards, they gave the Unix servers of the work NFS, NIS, the RPC system and a host of improvements to Unix, everyone else's Unix would have to have the lines copyright (c) Sun Microsystems, Inc. in various important places.
Also in the late 80s and early 90s, some things were happening elsewhere in the world, that at the time might have seemed minor but would grow so very huge they would threaten the life of Sun (and quite a few others in the IT business). Stallman and FSF with dream of free OS including compiler and toolset and support utilities all operating systems need, Linus with Linux. Stinky ol' segmented x86 and peripherals expanding to where they could do bigger jobs (even if still retaining some trashy heritage to this day from its past)
Somewhere along the line Sun and other Unix vendors stopped even including the compiler for free, after all, most people didn't use it and those who did could be made to pay thousands. Problem with that was there was now free tools out there from the FSF and other sources (like replacements for Unix utilities that were sometimes superior to the ones included with Unix, and things like web servers that corporations were finding useful just for internal use only even if they weren't on the Internet in a customer facing manner). But that was no big deal, after all the GNU compiler (we call gcc now) was free and available and so started to be used first for building those nifty free tools, then do fill in to do customizing since those vendor compilers were so very dang expensive, and pretty soon their were companies with internal applications and administration tools either fully written or built with gcc/toolset or on handy languages like Perl built with gcc.
During the dot-com boom Sun had huge demand and profit, but people were buying the Sun name and Sun slacked off innovating. Around 2002 GNU/Linux and free software could do most of what Unix on server machines could do, and if something could do 85% of the job on hardware that cost 40% as much, something upsetting to the picture was bound to happen. And developers writing free software and other types of open software could have robust and relatively unencumbered communities, while those in closed source world has to be done in a very careful and stifling manner with oversight by nervous proprietors (I've been developer, software engineer, and manager in closed source corporations, community there is a whole different ball game, and often is tied to trade shows, schmoozing customers, paid support, etc.)
Sun lost market share to GNU/Linux and the hardware it can run on. No surprise, GNU/Linux scales and runs on everything from cell phones to supercomputers, that's a kind of scalability and breadth of hardware platform and device support SunOS/Solaris and all other Unix never had and still doesn't have.
Eek, says Sun, what to do? Now Sun still has R&D pluck and spirit about it, for that and some other things I admire them even to this day. But the way they go at that can be awesome for their customers and investors, or just maddening. They'll sink tens of millions into a "trial balloon" which might get dropped and leave their customers with their metaphorical pants down. They've been into and out of and into again x86 solaris, for example. Into and out of and back into GNU/Linux support. Into and out of alternate architectures (Cobalt MIPS, for instance). Into and out of thin client push. Into enterprise storage, check back later.
So what's so funny in this article, which is really about Sun trying to do a FSF and Linuxy type thing maybe a decade and a half too late, a kind of "Woodstock II - build it and they will come", and maybe just as another one of many multi-million dollar trial balloons that could pop any second if Sun decides results weren't worth effort ?
First, there's the potshot at Linux by the author, implication it might have same source code as Unix (internals of Linux are nothing like Unix, a person who's not a coder but who has technical mindset can even find that much out just seeing block diagrams of internals of Unix in classic old texts.). Sounds like tired regurgitating of SCOs claims about Linux, which have been 99% dismissed as nonsense by our courts, and the future of the couple of remaining claims looking very bad indeed, (as SCOs stock prices after this pump and dump ploy), it's metaphorically much like a lingering fart in church after the congregation long left. Stephen might give some thought that the legal geniuses at IBM have already taken a long hard look long ago at LInux' guts, and, along with their marketing, sales, and tech jumped on the bandwagon with full commitment and confidence, and are hugely successful with it, and the claims of SCO are crumbling like toy bucket-built sand castles on the beach.
The next funny thing is that Sun would love to have Linux's device support, so much their thinking about "shims" to allow Linux written stuff to be used in the kernel, but of course Solaris isn't free software and almost all of those nifty device thingies are. And if Sun changes to a free software license, they'll be in danger of losing control of Solaris, and also of Linux developers merely ripping the kewl parts out of the warm bowels of Solaris to sew into Linux and once again leaving Sun with even faster dwindling market share.
Another haha is the mentioning of what I and all too many Solaris consultant/support/employee has to do every freakin' time I go to a huge corporate or government Sun box and log in, the set -o mumbledeefoos because Sun by default uses such a stinky old shell with piss poor support of whatever terminal/console type one has at the moment, that even the freakin backspace keys don't work (and don't even think about touching arrow keys). And no one wants to change the Sun defaults or install any of the other shells Solaris could run (and for craps sake don't change root's default shell). See, Sun doesn't have the number of developers to throw at those nice creature comforts as the GNU/Linux people do, never will, so they have to either get those folk working on Solaris problems or pay money to play catchup. GNU/Linux was passing up the commercial Unix in this regard over eight years ago, polish in terminal/console mode and polish in administration tools, and now they're just left behind.
Which brings us to how the usual Unix utilities work, they mention the directory listing command "ls", how that's different from the way the standard GNU/Linux one is. See, the big Unix vendors all have slight differences about how many common command work and barf stuff back at you out of the box, and it can get annoying for people like me who work on a Sun Solaris box one week, and AIX or HP/UX or IRIX or Unixware the next. Some of those have even evolved to putting data on different lines rather than separate columns, nevermind the values and units of what you get back. Oh sure, there's a way to get older standard behaviour, just let me fumble around in the man pages a few minutes and I'll get back to you with the variable(s) to set or in some cases the path to the flavor of the command (same name different output) you want. But this fragmentation has made the Unix vendors hurt each other, harder to change, harder to make universal scripts. In the meantime the GNU/Linux way (and to be fair the open source *BSDs also helped with this) became the de-facto standard - what a guy who wants to learn Unixy-way most likely is going to do is run GNU/Linux at home on his PC, and that is what becomes the benchmark the other Unix are increasingly being held to, because it's just better, more convenient and better thought out (project with a focus versus 30+ years of add-ons, improvements, experiments by committees at a dozen companies going in a dozen directions trying to all be different and trying to lock in their way of doing things to the exclusion of the others)
Speaking of polish, the install of commercial Unix might be OK if you go with the vendors' box and vendor's options, but it's hell with third party boards and devices. And adding a disk (or array of disks) or peripheral later can be fiendishly complicated on a commercial Unix, even can be hard to figure out what the new device's name is. I've actually written scripts for clients just so they can have hope in hell of finding right SCSI lun numbers and fibre HBA WWN out of a pile on Solaris and HP/UX (heh, or even which driver is being used for what) It's generally easier in the GNU/Linux world (must be, no one's ever called for help with that after I do Linux migration and show them the five minute device-finding spiel)
They want the minds (and manhours) of the Linux developers. Now what is a hoot is that most of those Linux folk want their hard work to be used far and wide everywhere and anywhere. Not on one particular vendor's box, and not under the main control of one company, and not tied to the future of one company. I'm even talking about the paid developers at RedHat and SuSE and Ubuntu and Mandrake, when they write something they know if its good and if it's useful it'll pop up in Debian or Arch or Slack, or it might get improved over there and put back. Not to mention it's already being used everywhere and anywhere whereas Solaris re-adoption by a majority of the market is still an open question.
My advice to Sun would be to get 100% on the GNU/Linux train, even to contributing fine grained locking and dtrace and other nifty tweaks that give Solaris an edge at the moment over Linux. Make compatibility libraries and shims for Solaris targeted software and drivers. Then the Linux developers and the application developers who target Linux firstly work for you. Solaris development can be tapered off, no reason to lose customers or the huge application base or compatibility, but grow the GNU/Linux and shrink the Solaris over the next five or more years. Because the hundreds of millions of dollars a year it takes to have your own Unix is getting too expensive in this fiercely competitive business, and Sun's competitors have already made the choice to do it and Sun is not as big as they are nor have their cash reserves.