Microsoft to Stop Releasing Services for Unix

lilrowdy18 writes "According to a recent article, Microsoft will stop releasing any new versions of Services for Unix. SFU 3.5 will continue to be supported until 2011 and will have extended support until 2014. From what the article hints at, Microsoft wants Unix interoperability integrated into the OS. Microsoft says that this integration couldn't be done with past architectures."

The full story

[mparaz]

The eWeek article is just a summary. The full story is here [microsoft-watch.com].

Re:Integrated

[b100dian]

You can use PAM LDAP for login against an AD.
And, of course, samba.

Is this enough "Client for Microsoft Networks"? :)

Not really discontinuing SFU...

[RhettLivingston]

but reinventing it. By moving this capability into the OS instead of hosting it as a parallel OS on the same kernel, they will gain performance and increase integration.

This is actually just one more example of an acceleration of rumors of Longhorn features. The rumors were that Longhorn would be able to run Unix applications and, specifically, x86 Linux binaries without recompilation. It looks now like at least a portion of that capability will appear in SP2 for Win 2003 Server a full year before Vista release.

Re:Microsoft's bait and switch

[Bluey]

Why pity them? They will still be supported for another 6 years (9 if they want extended support). They just aren't releasing any new versions of it.

Even still, some of the next-gen SFU functionality is being integrated [microsoft-watch.com] into Windows Server 2003 R2. It's not the end of unix interoperability from Microsoft, just this derivation of it.

Re:What happened to ten years?

[Bluey]

It was released in January of 2004, so mainstream support should end 2009, extended support ends 2014. Sounds like they decided to extend mainstream support 2 years to 2011 and still end extended support in 2014. No conspiracy to see here.

Re:Integrated

[Tony Hoyle]

If 'ls -l' is doing a lookup for each line then your nscd is not running or broken.

All user information (and host) on Unix is cached - and the cache is *not* a linear lookup.

Username/PAM lookup is *not* linear. If I call getpwnam for example it goes to pam -> active directory -> username lookup. There's no searching involved.

The trouble with cygwin

[petermgreen]

is its posix on win32 rather than posix on NT

This makes certain things (most notablly select) rather difficult to implement and slow.

Re:Integrated

[HairyCanary]

I have a gaggle of Solaris boxes that authenticate to LDAP (which AD is) and I do not see any appreciable delay due to the username lookups. And yes, our LDAP directory has thousands upon thousands of users.

Re:Not really discontinuing SFU...

[brajesh]

Exactly, as Bill Hilf, MS's linux lab manager pointed out [slashdot.org] on recently on /.

"I can confirm that the next-generation of several components of Services for Unix are being integrated into Windows Server 2003 R2. The Network File System (NFS) client, NFS Server, User/Name Mapping, Telnet Server & Client, Password Sync and NIS Server components of Services for Unix are all present in the Windows Server 2003 R2 builds[...]In addition, a revamped POSIX subsystem, the 'Subsystem for Unix-based Applications' or 'SUA' is also available as an optional install in R2."

Re:Integrated

[Dolda2000]

I don't know what Unix system you use, but on my GNU/Linux system (which uses the same APIs), based on GNU libc, the get*ent APIs are implemented using nameswitch modules, which can do lookups in LDAP, NIS, /etc/passwd, /etc/passwd.db, a MySQL database, or anything else. And indeed, it will be on the complexity order of whatever that algorithm chooses -- it's not a flat search.

I will agree that there are a lot of things that should be done with the Unix "directory services", but not that which you describe. The greatest problem is that Unix still uses numeric UIDs, whereas it should be using symbolic UIDs (such as Kerberos principles).

Never understood the name

[vijayiyer]

Shouldn't it have been called Unix Services for Windows? In another example of MS marketing spin, they act as if SFU somehow does something for Unix, when it instead adds basic functionality to Windows. I used SFU for about 1 month. I still was so frustrated doing Fortran development under Windows that I wiped the drive and installed Linux.

BillG claimed in 93 that NT was UNIX

[Anonymous Coward]

The following piece was in Communication Week printed in June/July 1993 (no 461, p.8):

UNIX IN NT'S CLOTHING

In his latest positioning statement on Microsoft Corp.'s Windows NT operating system, company chairman and CEO Bill Gates said NT is not a competitor of Unix, but in fact uses the same kernel. "I think [NT] will very quickly be the most popular form of Unix out there, because we do not allow licensees to change it around to try and get proprietary advantages on top of what was on there," Gates said at last week's PC Expo in New York. "NT is a form of Unix. It will not replace Unix, but I expect it to be the most popular form of Unix."

Re:Windows Vista to get POSIX subsystem

[sconeu]

NT4 actually had a (not very functional) POSIX subsystem. It was needed to get certain government contracts that specified POSIX. The POSIX subsystem was removed/deprecated in Win2K.

Re:Windows POSIX implementation

[sconeu]

Actually, the Santa Cruz Operation became Tarantella, which was recently bought by Sun.

The current entity calling themselves "The SCO Group" is what used to be called Caldera. They bought *something* from Santa Cruz (definitely their Operating Systems Division) and some sort of assets (but they can't produce the purchase agreement), and changed their name from Caldera to SCO. Allegedly this was for name recognition/branding, but apparently was really to sow confusion for their lawsuits.

Re:Kill Interoperability?

[RupW]

I believe it was Interix (not InterOp as sibling post has it).

Interix was the product name, which survives today as part of SFU 3.5. The original vendor was Softway Systems. They wrote Interix; Microsoft bought Softway and rolled Interix 2.2SP1 into SFU 2.0.

The confusion, I guess, is that InterOp systems bought the domain interix.com. They now sell util ports to interix and interix-related services. But AFAIK InterOp had nothing to do with SFU/Interix itself.

Re:The full story

[barking_at_airplanes]

Hardly shocking information since it was announced back on April, 26, 2005 in the Interix Forums (http://www.interopsystems.com/tools/forum [interopsystems.com]). So this has been known *publicly* for over 4 months now with a lot more detail than is in the other articles. The link is http://www.interopsystems.com/tools/forum/tm.aspx? m=5623 [interopsystems.com]

The real story is not that SFU is ending, but rather that it is becoming part of the regular distribution. Since January, 2004 SFU version 3.5 has been available as a free download from Microsoft (http://www.microsoft.com/windows/sfu [microsoft.com]) with no restrictions, etc. So this change has been expected for some time.

Complaints in other posting about "missing" applications are pretty hollow: two compilers are available (gcc 3.3 and MSVC {a free version of MSVC is available}, OpenSSH, bash, zsh, etc. They can be picked up freely from the Interix Tools site http://www.interopsystems.com/tools [interopsystems.com].

Re:Integrated

[Just Some Guy]

All user information (and host) on Unix is cached

That's true, where "Unix" == "Linux with nscd installed and running". Don't feel bad, these kinds of assumptions aren't new [retrologic.com].

Re:Microsoft's answer to UNIX

[dar]

Your statement doesn't make any sense. Win32 is an API of which NT is one implementation.