Sun Says Project Indiana is Not a Linux Copy

eldavojohn writes "Ian Murdock (Debian author & Sun's OS Chief) made some comments about Project Indiana that many have said is an attempt to make Solaris simply "more Linux-like." But Murdock quashes any concerns that this is just another Linux clone — muddying up the waters of distribution selection. He says that it's more a 'best of both worlds' attempt to make an OS that appeals to a broader audience. From the article, "Project Indiana will include a revamped package management system, which should prove popular with developers unaccustomed to Solaris. The OS has some clunky, archaic aspects, and Murdock thinks the new package system will modernize Solaris.""

Comments

Re:Bummer...

[wellingj]

Yes you are indeed a troll. But mostly because you are talking out your ass.
In 1996, Bruce Perens replaced Ian Murdock as the project leader.

package management

[Peter La Casse]

It would be great to see Solaris become tightly integrated with something like apt. pkg-get is ok, but it isn't currently used for all packages, and a Sun-backed and -improved version would be better. For example, I'd like to see it manage security updates in a way that meets the needs of Solaris sysadmins, with separate actions for downloading, applying and rolling back. I'd also like to see my attempts to install gvim not download 50 megabytes worth of libraries that are already on my system, in a slightly different version number.

Good Gnus?

[otis wildflower]

Beyond fixing software distribution and pkg mgmt (which is lonnnnngggg overdue!!), how about making GNU utils the default and tossing the archaic Solaris versions of common tools into some compat directory? If the GNU tool doesn't support some Solarisism (like, say, RBAC or extended attributes), hack the GNU tool and release the change as GPL.

Oh, and while you're refactoring, please fix JES. It is a clusterfuck mess, particularly the Delegated Administrator.

Re:Good Gnus?

[AaronW]

I remember reading a discussion by the Opensolaris folks claiming how much better their tools were than the Gnu tools. They were singling out Solaris TAR and saying what a mess the Gnu version is. As a user, Solaris TAR is crap to use compared to GnuTAR. It can't even handle tar files over 2GB, for example, and I've had several tar files it can't handle that GnuTAR can. I also like the built-in support for bzip2 and gzip in GnuTAR. Now granted, I haven't tried Solaris 10, but I suspect the problems remain. That isn't to say that GnuTAR is perfect... I was running an older version that would truncate some filenames.

Another one that caused me endless frustration was Solaris newgrp did not allow you to specify a command line like the Linux one did. I ended up porting over the Gnu version just so I could do my job without having to manually type in a command as part of a build procedure.

At my job I've been maintaining KDE for Solaris for a bunch of Sun users. When I migrated my desktop to Linux I'm still having to support KDE for them (and that's not my job, I'm not in IT). None of the developers like Sun's Gnome 2.0 that's included with Solaris 9, and newer versions have problems compiling because Sun still does not support the X render extension (on Sparc). Trying to compile KDE was difficult, since Sun's libraries and tools are often broken or so out of date. I also have had to compile GCC for Solaris to do it, since Sun's C++ compiler we have barfs on a lot of open source packages. I've also hit a number of problems because numerous features are missing in Sun's libc, even though they've been part of the posix or ISO standard for many years (i.e. missing stdint.h), including some parts of stdio.h. (stdint.h is part of the ISO C99 standard, well before Solaris 9 came out).

I remember having tons of problems with Sun's sed because it would silently truncate after 4-8KB of input.

-Aaron

Re:Bummer...

[Anonymous Coward]

Mod me troll if you must

Good idea. I'm almost tempted to give you an "insightful" for that suggestion, but it'd rather detract from the "troll" rating.

Re:solaris is starting to sound good

[wiredlogic]

Solaris is a pretty nice system overall. Sun's biggest failing from a user experience is their adherence to obsolete versions of the standard *NIX applications. Most of the stuff in /bin has none of the useful features added by POSIX. The POSIX stuff is all sequestered in /usr/xpg4/bin. This is a PITA when you want to write portable shell scripts that aren't restricted to a 25-year old subset of UNIX.

Re:Pithy Aphorism: "If you cannot beat them ..."

[kv9]

Otherwise, it's a toy OS, just like that Microsoft thing. what, Xenix?

The sincererest form of flattery

[Zigurd]

Sun has been groping for a way to compete with Microsoft for over 10 years. Well, "groping" might be too harsh, considering the strategy consisted mainly of denial about the fact that Windows on commodity hardware could run serious applications.

Ubuntu showed the way in both how to do it and the right business model, and Sun has done absolutely the right thing by directly imitating the Ubuntu way by becoming, effectively, a downstream Debian distro. Heck, they hired Ian Murdock to make sure you get it right. At Sun, this is probably necessary because corporate conservatism about cannibalizing revenues would have watered down a purely internal initiative.

Sun could still screw it up. There are plenty of weasel words like "two tier" in this article. But if Sun gets it right and "dissolves" Solaris into a number of userland projects and a kernel alternative to Linux (the way GNU Hurd theoretically is), and executes an a la carte support model like Canonical, they deserve to win a big slice of the business.

Re:Sun as usual is copying IBM

[fimbulvetr]

I dont understand how Sun can be seen as innovative anymore. They just lurch this way and that, never following any kind of coherant strategy.

No need to try to reverse engineer their strategy, it's openly published:

http://media.arstechnica.com/journals/microsoft.me dia/sunstrategy1.gif

Re:solaris is starting to sound good

[AaronW]

Exactly. I can't count how many scripts fail on Solaris 9 because of Sun's /bin/sh missing some key functionality (usually replacing it with /bin/bash fixes it). And why should scripts have to hunt around all over the place just to find a working version of very common tools (like Sun's sed which used to be quite broken). And some very useful features are always missing (recursive grep anyone).

Trying to compile GNU software on Solaris 9 is often a painful experience because even their libc and header files are in the dark ages (i.e. many ISO C99 features are missing). I haven't tried Solaris 10 and moved on to Linux at work.

Re:Why look at Solaris now?

[Anonymous Coward]

- Much finer grain security (RBAC principles in Solaris are much better handled than in grsecurity for GNU/Linux).

- ZFS is now bootable (ZFS is now bootable)

- DTrace is much more powerful than strace (a number I read in one of the Sun DTrace presentation stated ~40000 probe points in Solaris to ~40 in GNU/Linux).

- All the p* tools on Solaris are much more powerful (and in some instances, there are no equivalents in GNU/Linux requiring users to have to code their own inotify CLI program)

- Service management is also much better taken care of with SMF than anything else I've seen in any of the BSDs or GNU/Linux as of late.

- Project Athena (used for handling massive users+groups / computers invented by MIT) runs with SunOS (and I believe Solaris as well))

- All GNU tools are available in /usr/sfw, all BSD tools in /usr/ucb, plus all of Solaris' goodness.

- Unlike GNU/Linux, Solaris adheres more strictly to standards much better and so if you have to write programs that are to be cross platform... developing them on Solaris will probably be your best bet of seeing if it works on other platforms.

- Another thing (that is if this matters to you), it's the only open source Sys V UNIX available.