From: Eric Lee Green <e...@prysm.net>
Subject: Re: Window NT vs x86 solaris vs SCO?
Date: 1997/03/20
Message-ID: <33320A3F.77E80BA6@prysm.net>#1/1
X-Deja-AN: 227165462
References: <4ucul0$qav@knot.queensu.ca> <DvzHyn.Hz1@telly.org> 
<4umfiu$9nh@hobyah.cc.uq.oz.au>
Organization: Executive Consultants
Newsgroups: comp.unix.solaris,comp.unix.sco.programmer,comp.unix.sco.misc,
comp.os.ms-windows.nt.advocacy,comp.os.ms-windows.nt.misc


John Wiltshire wrote:
> Note you said "all the internet software".  There is a lot more
> software in the world than Linux software, especially if the company
> already has an installed base of PC systems, NT is a fairly good
> choice.

Err, you use Linux as the server (has a pretty good SMB implementation,
and there does exist
a full-featured DHCP implementation, though the Red Hat 4.1 release
mysteriously left it out).
You run your software on your Windows 95 or Windows NT Workstation
clients. Works like a Novell server, i.e., you don't run your software
on the server, you run it on the client. (The exception, of course, is
database servers, of which there are many good NT and Linux/Unix ones,
and few that run on the Novell side of things). 

Linux Advantage: Lower costs, higher performance.
Disadvantages: More difficult to configure. Have to reboot it about 4
times a year to upgrade your OS over the Internet or from a new CDROM
(that's how Linux vendors make their money, by releasing new releases at
supersonic pace and hoping you'll buy their new CD because they quit
supporting the old release). 

SCO Unix also has a pretty good SMB implementation. I've been running
the Free SCO here. The latest SCO 5 release is pretty slick, and quite
as easy to use and configure as your typical NT box (much slicker than
the Linux, though Linux has gotten much better over the past year). 
My only beef with SCO Unix is that it is so slow at accessing EIDE
drives -- you pretty much have to go SCSI with it to get decent hard
drive speeds (as vs. Linux or NT, both of which will do full 32-bit
bus-mastering DMA multi-sector access with modern chipsets).  There's
also the fact that you have to tune the kernel to get the best
performance :-(  (Linux dynamically adjusts according to the load, while
NT -- well, NT doesn't bother with tuning, NT expects you to buy a
faster computer :-). 


> NT is good where a high end Unix system would be overkill as a general
> rule.  Many people are moving towards NT because of its tight
> integration with Win 3.x and Win95 - Unix servers don't have this
> integration and are unlikely to see the need for it.

Err, you have been away from the Unix world for quite some time. I
regularly access Win95 shares from my Linux box and have Win95 machines
access some of my directories as shares (in particular, that makes a
VERY slick IntraNet FTP server -- just point your Network Neighborhood
at my FTP directory, and use point-and-click!). I don't think SCO Unix
can mount Windows file shares yet, but it certainly will print to
Windows print shares with no problem and has no problem exporting
directories as WIndows shares. 

> For high end apps, NT can't cut it yet.  For a cheap upgrade in a
> Windows dominated network, NT is a good choice.

My take on it: Right now, the best reason to go NT is because it's much
easier to administer than Novell. The traditional hack on Unix is that
it's hard to administer. That, however, has been changing, due to
GUI-based administration tools such as provided by SCO and Red Hat
Software. 

> >Solaris makes a good product in Solaris X86, but the company is in a
> >quandry. Hardware sales are far more lucrative, and if they make
> >Solaris X86 *too* good it'll cut into SPARC sales. That is something
> >Sun would dearly want to avoid. SCO and Linux developers, since they
> >don't sell hardware, have no such artificial restrictions.

The outfit I work for looked into selling Solaris X86, but Sun was more
interested in selling
us Netras (SPARC) and I couldn't even find someone willing to quote us a
price on the X86 product. The educational administration group (our
area) was quick to say it was available, but was even quicker to try to
sell us a Netra. Sorry, we're software vendors, not hardware vendors
:-(. 

--
Eric Lee Green e...@prysm.com

From: j...@qits.net.au.nospam (John Wiltshire)
Subject: Re: Window NT vs x86 solaris vs SCO?
Date: 1997/03/27
Message-ID: <333a962b.8454847@news.uq.edu.au>
X-Deja-AN: 228761838
References: <4ucul0$qav@knot.queensu.ca> <DvzHyn.Hz1@telly.org> 
<4umfiu$9nh@hobyah.cc.uq.oz.au> <33320A3F.77E80BA6@prysm.net>
Organization: Traffic Systems Development
Newsgroups: comp.unix.solaris,comp.unix.sco.programmer,comp.unix.sco.misc,
comp.os.ms-windows.nt.advocacy,comp.os.ms-windows.nt.misc


On Thu, 20 Mar 1997 22:10:39 -0600, Eric Lee Green <e...@prysm.net>
wrote in comp.os.ms-windows.nt.advocacy:

>John Wiltshire wrote:
>> Note you said "all the internet software".  There is a lot more
>> software in the world than Linux software, especially if the company
>> already has an installed base of PC systems, NT is a fairly good
>> choice.
>
>Err, you use Linux as the server (has a pretty good SMB implementation,
>and there does exist
>a full-featured DHCP implementation, though the Red Hat 4.1 release
>mysteriously left it out).
>You run your software on your Windows 95 or Windows NT Workstation
>clients. Works like a Novell server, i.e., you don't run your software
>on the server, you run it on the client. (The exception, of course, is
>database servers, of which there are many good NT and Linux/Unix ones,
>and few that run on the Novell side of things). 

I dispute that there are "many good Linux DB servers".  Unix perhaps.

>Linux Advantage: Lower costs, higher performance.
>Disadvantages: More difficult to configure. Have to reboot it about 4
>times a year to upgrade your OS over the Internet or from a new CDROM
>(that's how Linux vendors make their money, by releasing new releases at
>supersonic pace and hoping you'll buy their new CD because they quit
>supporting the old release). 

I think you'll get an argument about the rebooting.

You also forgot the lack of fault tolerant drive systems, but who's
arguing?

>SCO Unix also has a pretty good SMB implementation. I've been running
>the Free SCO here. The latest SCO 5 release is pretty slick, and quite
>as easy to use and configure as your typical NT box (much slicker than
>the Linux, though Linux has gotten much better over the past year). 
>My only beef with SCO Unix is that it is so slow at accessing EIDE
>drives -- you pretty much have to go SCSI with it to get decent hard
>drive speeds (as vs. Linux or NT, both of which will do full 32-bit
>bus-mastering DMA multi-sector access with modern chipsets).  There's
>also the fact that you have to tune the kernel to get the best
>performance :-(  (Linux dynamically adjusts according to the load, while
>NT -- well, NT doesn't bother with tuning, NT expects you to buy a
>faster computer :-). 

You don't look at NT much, do you?  NT dynamically adjusts to load as
well and you can tune the kernel through the registry as much as you
like.

>> NT is good where a high end Unix system would be overkill as a general
>> rule.  Many people are moving towards NT because of its tight
>> integration with Win 3.x and Win95 - Unix servers don't have this
>> integration and are unlikely to see the need for it.
>
>Err, you have been away from the Unix world for quite some time. I
>regularly access Win95 shares from my Linux box and have Win95 machines
>access some of my directories as shares (in particular, that makes a
>VERY slick IntraNet FTP server -- just point your Network Neighborhood
>at my FTP directory, and use point-and-click!). I don't think SCO Unix
>can mount Windows file shares yet, but it certainly will print to
>Windows print shares with no problem and has no problem exporting
>directories as WIndows shares. 

So, how can I get a single network logon to all file servers,
databases, email, intranet (as in the NT domain model) through the
simple SMB stuff on Linux.  Maybe you haven't looked at NT for a while
but there is a lot more to networking than the SMB, SMB, SMB that you
keep falling back on.  NT does SMB a lot better than Linux (primarily
because MS won't release all the specs).

>> For high end apps, NT can't cut it yet.  For a cheap upgrade in a
>> Windows dominated network, NT is a good choice.
>
>My take on it: Right now, the best reason to go NT is because it's much
>easier to administer than Novell. The traditional hack on Unix is that
>it's hard to administer. That, however, has been changing, due to
>GUI-based administration tools such as provided by SCO and Red Hat
>Software. 

Administration of any system is a long process to do well.  There is
nothing more dangerous than someone who thinks NT is easy to admin -
you'll be looking at an insecure system in no time flat!

>> >Solaris makes a good product in Solaris X86, but the company is in a
>> >quandry. Hardware sales are far more lucrative, and if they make
>> >Solaris X86 *too* good it'll cut into SPARC sales. That is something
>> >Sun would dearly want to avoid. SCO and Linux developers, since they
>> >don't sell hardware, have no such artificial restrictions.
>
>The outfit I work for looked into selling Solaris X86, but Sun was more
>interested in selling
>us Netras (SPARC) and I couldn't even find someone willing to quote us a
>price on the X86 product. The educational administration group (our
>area) was quick to say it was available, but was even quicker to try to
>sell us a Netra. Sorry, we're software vendors, not hardware vendors

We found a similar thing when we were looking for firewall products.
Sun kept wanting to give us Netras with Firewall-1 when we had a stack
of 486 PCs lying around that would have done fine with Solaris x86 and
the same product.  We ended up just forgetting the whole thing and
going with Borderware.

John Wiltshire

From: j...@cs.toronto.edu (John DiMarco)
Subject: Re: Window NT vs x86 solaris vs SCO?
Date: 1997/04/09
Message-ID: <1997Apr9.174306.8063@jarvis.cs.toronto.edu>#1/1
X-Deja-AN: 231967386
X-Nntp-Posting-Host: gardiner.cs.toronto.edu
References: <4ucul0$qav@knot.queensu.ca> <33320A3F.77E80BA6@prysm.net> 
<333a962b.8454847@news.uq.edu.au> <slrn5k3irg.27e.champ@44mag.vistech.net> 
<E8DLnJ.CtI@bigbird.telly.org>
Newsgroups: comp.unix.solaris,comp.os.ms-windows.nt.misc,comp.unix.sco.programmer,
comp.unix.sco.misc,comp.os.ms-windows.nt.advocacy


>>In article <333a962b.8454...@news.uq.edu.au>, John Wiltshire wrote:

>>>but there is a lot more to networking than the SMB, SMB, SMB that you
>>>keep falling back on.  NT does SMB a lot better than Linux (primarily
>>>because MS won't release all the specs).

Can someone please provide some evidence that NT does SMB "a lot better" than
Linux, by some reasonable definition of better?  I'd be interested in seeing
some data to substantiate or refute this claim.

Regards,

John
--
John DiMarco <j...@cs.toronto.edu>                         Office: SF2101
CSLab Systems Manager                                     Phone: 416-978-5300
University of Toronto                                     Fax:   416-978-4765
http://www.cs.toronto.edu/~jdd

From: dlebl...@mindspring.com (David LeBlanc)
Subject: Re: Window NT vs x86 solaris vs SCO?
Date: 1997/04/10
Message-ID: <33503aad.2104291721@news.mindspring.com>#1/1
X-Deja-AN: 231929740
Distribution: inet
References: <4ucul0$qav@knot.queensu.ca> <33320A3F.77E80BA6@prysm.net> 
<333a962b.8454847@news.uq.edu.au> <slrn5k3irg.27e.champ@44mag.vistech.net> 
<E8DLnJ.CtI@bigbird.telly.org> <1997Apr9.174306.8063@jarvis.cs.toronto.edu>
X-Server-Date: 10 Apr 1997 01:03:48 GMT
Organization: MindSpring Enterprises
Newsgroups: comp.unix.solaris,comp.os.ms-windows.nt.misc,comp.unix.sco.programmer,
comp.unix.sco.misc,comp.os.ms-windows.nt.advocacy


j...@cs.toronto.edu (John DiMarco) wrote:

>>>In article <333a962b.8454...@news.uq.edu.au>, John Wiltshire wrote:
 
>>>>but there is a lot more to networking than the SMB, SMB, SMB that you
>>>>keep falling back on.  NT does SMB a lot better than Linux (primarily
>>>>because MS won't release all the specs).
 
>Can someone please provide some evidence that NT does SMB "a lot better" than
>Linux, by some reasonable definition of better?  I'd be interested in seeing
>some data to substantiate or refute this claim.

For one thing, I can set permissions on my NT files across the
network.  I'm also able to manage shares, etc. very easily.  Although
it seems to have gotten a lot better over the last year, I've also
seen samba servers crash Win95 (no great loss, but the user was
annoyed...).  This one is fixed, but at one time if you copied a
read-only file to a samba server, it was stuck - no way to delete it
or change the flags except from the UNIX CLI.

I've also seen situations where bugs in samba would feed NT stuff it
didn't expect and crash it (bad code on both sides).  Hopefully, most
of this will abate as CIFS becomes a recognized and developed
standard.

I've also had Linux keep my 2 domain controllers from talking properly
- the Linux box would claim to be a PDC just well enough to screw up
the BDC.  John is right that if everything were more open we wouldn't
have so many of these problems.


David LeBlanc           |Why would you want to have your desktop user, 
dlebl...@mindspring.com |your mere mortals, messing around with a 32-bit 
                        |minicomputer-class computing environment?
                        |Scott McNealy

From: Jeremy Allison <j...@cygnus.com>
Subject: Re: Window NT vs x86 solaris vs SCO?
Date: 1997/04/10
Message-ID: <334D8633.735F@cygnus.com>#1/1
X-Deja-AN: 232175778
References: <4ucul0$qav@knot.queensu.ca> <33320A3F.77E80BA6@prysm.net> 
<333a962b.8454847@news.uq.edu.au> <slrn5k3irg.27e.champ@44mag.vistech.net> 
<E8DLnJ.CtI@bigbird.telly.org> <1997Apr9.174306.8063@jarvis.cs.toronto.edu> 
<33503aad.2104291721@news.mindspring.com>
Organization: Cygnus Solutions
Reply-To: j...@cygnus.com
Newsgroups: comp.unix.solaris,comp.os.ms-windows.nt.misc,comp.unix.sco.programmer,
comp.unix.sco.misc,comp.os.ms-windows.nt.advocacy


David LeBlanc wrote:

> I've also seen situations where bugs in samba would feed NT stuff it
> didn't expect and crash it (bad code on both sides).  Hopefully, most
> of this will abate as CIFS becomes a recognized and developed
> standard.

No David I'm not letting you get away with that. If a server crashes
due to something the client sends it is is a *SERVER* bug. Period.

You should be able to open a connection to an NT or Samba server
cat a core file down there and the server shouldn't crash. If it 
does it's a *SERVER* BUG !!!!!!!!!!

Or do you just enjoy repeating what Microsoft posted on their
Web site at the time - "Samba client sends 'bad' commands".

Bad commands, naughty commands, go and stand in the corner :-).

Now I'm not saying Samba doesn't have similar bugs, but when
we do we admit to them and we don't blame the MS client. When
coding Samba I have to assume that anything an MS client does
is 'right' by default, as the CIFS spec isn't good enough yet.

Also, it wouldn't even matter if we were right and they were
wrong - people expect their MS clients to work with Samba -
that means what their clients do *define* the standard, not what
the spec says, unfortunately.

> For one thing, I can set permissions on my NT files across the
> network.  

Undocumented by MS. Care to spec it for us ? Do you think
they would if you asked them ? Try it sometime.

> I'm also able to manage shares, etc. very easily.  

Undocumented by MS. Care to spec it for us ? Do you think
they would if you asked them ? Try it sometime.

> Although
> it seems to have gotten a lot better over the last year, I've also
> seen samba servers crash Win95 (no great loss, but the user was
> annoyed...).  This one is fixed, but at one time if you copied a
> read-only file to a samba server, it was stuck - no way to delete it
> or change the flags except from the UNIX CLI.

Well we do try and fix all known bugs :-).

> I've also had Linux keep my 2 domain controllers from talking properly
> - the Linux box would claim to be a PDC just well enough to screw up
> the BDC.  John is right that if everything were more open we wouldn't
> have so many of these problems.

Then help us to pressure them to *DOCUMENT* it. Not that they would
listen to you any more than they've listened to us (and I've been
asking them for several years now).

Jeremy Allison.
Samba Team.

From: dlebl...@mindspring.com (David LeBlanc)
Subject: Re: Window NT vs x86 solaris vs SCO?
Date: 1997/04/11
Message-ID: <3350bc92.2203081343@news.mindspring.com>#1/1
X-Deja-AN: 232216593
Distribution: inet
References: <4ucul0$qav@knot.queensu.ca> <33320A3F.77E80BA6@prysm.net> 
<333a962b.8454847@news.uq.edu.au> <slrn5k3irg.27e.champ@44mag.vistech.net> 
<E8DLnJ.CtI@bigbird.telly.org> <1997Apr9.174306.8063@jarvis.cs.toronto.edu> 
<33503aad.2104291721@news.mindspring.com> <334D8633.735F@cygnus.com>
X-Server-Date: 11 Apr 1997 04:40:33 GMT
Organization: MindSpring Enterprises
Newsgroups: comp.unix.solaris,comp.os.ms-windows.nt.misc,comp.unix.sco.programmer,
comp.unix.sco.misc,comp.os.ms-windows.nt.advocacy


Jeremy Allison <j...@cygnus.com> wrote:

>David LeBlanc wrote:
 
>> I've also seen situations where bugs in samba would feed NT stuff it
>> didn't expect and crash it (bad code on both sides).  Hopefully, most
>> of this will abate as CIFS becomes a recognized and developed
>> standard.
 
>No David I'm not letting you get away with that. If a server crashes
>due to something the client sends it is is a *SERVER* bug. Period.

I'm in the middle of debugging something along these lines right now.
I said bad code on BOTH sides, and I meant it.  The samba client is
sending the NT box all sorts of hideously screwed up packets, and the
NT box keeps TRYING to tell it to go away.  It is certainly the fault
of the server if it dies from being sent rubbish, no matter how
hideous, but it is also the fault of the client for sending junk.

>You should be able to open a connection to an NT or Samba server
>cat a core file down there and the server shouldn't crash. If it 
>does it's a *SERVER* BUG !!!!!!!!!!

I agree.  However, if the samba client is sending something resembling
a core file (or actually much worse - something which masquerades as
valid SMB packets), then the client is also hideously screwed up.

Bad code on BOTH sides.  Client's fault for sending crap, and server's
fault for keeling over.

Take this to e-mail, and I'll explain in full - maybe you can help.

>Or do you just enjoy repeating what Microsoft posted on their
>Web site at the time - "Samba client sends 'bad' commands".

No, I'm SEEING it.  The samba client IS sending garbage.

>Bad commands, naughty commands, go and stand in the corner :-).

Right - shouldn't be sending garbage.  Server shouldn't be barfing on
it, either.  As I said - fault on both sides.

>Now I'm not saying Samba doesn't have similar bugs, but when
>we do we admit to them and we don't blame the MS client. When
>coding Samba I have to assume that anything an MS client does
>is 'right' by default, as the CIFS spec isn't good enough yet.

What do you call it when the client goes nuts and keeps sending
commands after the server tells it that there is an error and to
bugger off?  I call this a bad client.  When the server eventually
dies from abuse, I call this a bad server.
 
>Also, it wouldn't even matter if we were right and they were
>wrong - people expect their MS clients to work with Samba -
>that means what their clients do *define* the standard, not what
>the spec says, unfortunately.

Neither here nor there - when the server says "error - go away", the
client should dutifully drop the connection and quit annoying the poor
server instead of badgering it to death.

>> For one thing, I can set permissions on my NT files across the
>> network.  
 
>Undocumented by MS. Care to spec it for us ? Do you think
>they would if you asked them ? Try it sometime.

I _didn't_ say that you were at fault over this.  It is the case,
however.  Not saying it is fair, right, or proper.  Just that it is.

>> I'm also able to manage shares, etc. very easily.  
>
>Undocumented by MS. Care to spec it for us ? Do you think
>they would if you asked them ? Try it sometime.

See above.

>> Although
>> it seems to have gotten a lot better over the last year, I've also
>> seen samba servers crash Win95 (no great loss, but the user was
>> annoyed...).  This one is fixed, but at one time if you copied a
>> read-only file to a samba server, it was stuck - no way to delete it
>> or change the flags except from the UNIX CLI.
 
>Well we do try and fix all known bugs :-).

It does keep getting better.  I _like_ samba.  I think you guys are
doing a GREAT job without proper documentation, source, etc.

>> I've also had Linux keep my 2 domain controllers from talking properly
>> - the Linux box would claim to be a PDC just well enough to screw up
>> the BDC.  John is right that if everything were more open we wouldn't
>> have so many of these problems.
 
>Then help us to pressure them to *DOCUMENT* it. Not that they would
>listen to you any more than they've listened to us (and I've been
>asking them for several years now).

I try.  I just have to do what I can.  What you really need is someone
with fat pockets to buy you a source license.

BTW, thanks for the auditing tool - that was nice.


David LeBlanc           |Why would you want to have your desktop user, 
dlebl...@mindspring.com |your mere mortals, messing around with a 32-bit 
                        |minicomputer-class computing environment?
                        |Scott McNealy