Frank Cusack wrote:

> SMF/svc* is not an imitation of any Microsoft technology.

It's implementation is in "good" spirit of Microsoft tehnology IMO.

> It is a
> vast improvement over the old SYSV init system. All other systems
> seem quite antiquated in comparison.

It has some good ideas and features, but the current implementation is
ridiculous with it's binary database and vital system commands depending
on in (like routeadm(1m) in 8/07).

> I fail to see how anyone who
> looks at it in depth cannot agree.

Argumentum ad populum?

> Folks over 50 excluded (set in your ways). j/k

This is sort of ideological blackmail (let's for a moment abstract your
"just kidding" backout position). Not every new tehnology, or for that
matter "thing" in the world isn't good just because it is new. Having
slogans like "tehnology of 21 century", "modern approach" and so on are
often just hiding doubtful implementation or poor design of some thing
in IT. This appeal to someone's years, and this clasification of critics
like old horses (non intuitive, lost enthusiasm, gabbling, always
against new things and young bright ideas) is rethoricaly poor, but very
popular and efective argument today in "modern" times, where spot light
is effectively and de facto moved from criticised thing to critic itself
in order to shoot him with eggs and tomateos.

That said, I can say this: I have been for a 10 years now, and always
will be 'old'. I hate and despise GNOME desease, XML in config files,
"system tray", "desktop", "user friendly", "easy to use" and similar
senseless "modernisations".

Now, that said, I want to remind how I have fairly concrete complaints
and criticisms of some aspects of smf(5) - not an /arguments/ in a style
"this shit just sux" or "you are old".

BTW, I'm 0x20.


--
Miroslav

Richard B. Gilbert wrote:

> I used to work, as a contractor, for Mobil Research Laboratories ca.
> 1994 in Pennington, NJ. The desktop "standard" was Windows 3.1. If you
> called the helpdesk, somebody picked up the phone and said "Reboot!" If
> the basic first aid of rebooting didn't fix your problem, then and only
> then would they talk to you.
>
> We got our work done somehow or other but I can't say that Windows was
> much help in those days. W/2K was reasonably stable and a huge
> improvement over W/98. W/XP SP2 is still not perfect but I can go
> several days between reboots.

Of course, that's windows, not Linux.

BTW, Solaris kernel _is_ really stable, but uptime of regulary (internet
accesible) Solaris boxes is newer larger that month or so in contrast
with for example RHEL.

Why? Because every 3rd or 4th patch is patching some kernel driver or
part of the kernel. Patch and package maintainers are just labeling this
patches as "reconfigure", "reboot immediate", "reboot recomanded", not
bothering to check in postinstal or some post patch script if this
current machine is actually using that particular driver (modinfo) or
not. Not to mention patches which are patching /usr/bin/login (an
example) but are labeled as "reboot".

Of course, experianced sysadmin after some experiance with the system
can decide when to patch something and when is reboot really necessery,
but even than, reboots are most often than on other (similar) systems.

Of course, many background database servers and trusted intranet
machines are not patched so often, in mission critical oracle based
billing systems are never patched if there is no explicit and clear DBA
requests for that.


--
Miroslav

Thommy M. wrote:

> Hey, hey, hey, even us around 50 do love SMF. ;-)

You can pretend to love it, but you are getting old Thommy. You will be
replaced with young, more perspective forses soon. >:-)


--
Miroslav

Tim Bradshaw wrote:

> Tragically, this is true. I pretty much make my living patching
> machines, and Solaris 10 is just a complete horror compared to earlier
> versions.

My Linux colleagues are often laughing at me "it was more stable before
you patched it". Sad.

> I live in hope that Sun are learning their lessons about
> this, but I'm not sure anyone is really listening (and someone will
> say that this is because people like me aren't giving enough feedback,
> which is probably true).

Yes. Mr. Carlson said in couple occations here how people has to be more
active on mailing lists and forums, and not just lamenting here on
USENet. He is right. I will see if I can post some of my ideas and notes
someware else apart from this NG.

> Partly, of course, it's because Solaris 10
> is just harder to patch (what do you do in the presence of zones?),

I think this is not connnected with poor or missing checking scripts in
patches and errors in replaced ELF binaries themselves which are fixing
one thing, and breaking the other one.

> but in other ways it's because of bad decisions (introducing ZFS via a
> patch: just kill whoever thought that was a good idea,

This is not so bad as having for example SUNWnvsata empty package in
8/07 instead in some major update where people can downgrade whole
system update (or switch live update partition) if anything goes wrong.
In this way, some day maybe, there will be patch which will "populate"
SUNWnvsata package and replace cmdk(7d) as controller driver. I don't
know how many ck804 based chipsets will simply not work initially
because of small differencies and lack of testing, and machines will not
mount even root fs (boot from cdrom, chroot, patchrm), how will SVM
database, and /etc/lvm/* files get updated and so on ... Scarry.


--
Miroslav

Frank Cusack wrote:

> But really I was trying to question the earlier poster's judgement.
> svc > SYSV init > BSD rc,

But you failed to question it. Where are the tehnical arguments?

> and so obviously better that the only reason
> you wouldn't think so (well, IMHO) is if you either barely know svc
> (evidenced by comparison to Microsoft),

If we not look binary in the terms sux/rocks, it has some good features
and even fundamental idea.

> or if you have old-timer
> syndrome, like when folks thought the BSD rc.conf (etc.) system was
> better than SYSV init.

In the aspect of editing service properties, rc.conf as a text file is
more reliable than binary database which can get corrupted or simply
refuse to appy user changes with svccfg.

> -frank
> ps. I used 50 b/c a friend of mine who i've been trying to convince
> that IM and SMS are useful tools recently told me sorry he can't learn
> to operate/think that way,

He is right in SMS case. SMS is PITA to type on mob without proper user
friendly interface like keyboard.

> it's something that happens when you turn 50.

Even earlier in my case.

> of course he was kidding also, but you get the idea. But
> obviously, these days, 50 isn't old. 51 though, forget it ...

Of course, we are all kidding all the time ...


--
Miroslav

On 2007-11-18, Rich Teer <rich.teer@rite-group.com> wrote:
> On Sun, 18 Nov 2007, abuse@127.0.0.1 wrote:
>
>> While I agree it's better than sysvinit in principle, in practice
>> the only selling point is ability to restart dying daemons, which
>
> I disagree. The ability to start multiple services in parallel is
> useful too,

Well, maybe on a "can run 48 threads at once but only at 200MHz"
Sun box. How long do your services take to startup so that parallel
vs serial makes noticeable difference?

> as is the concept of dependancies.

That I agree with.

Dima

On Nov 19, 4:11 pm, ab...@127.0.0.1 wrote:

>
> Well, maybe on a "can run 48 threads at once but only at 200MHz"
> Sun box. How long do your services take to startup so that parallel
> vs serial makes noticeable difference?

Some services can take a long time to start (and no, not because they
are CPU intensive tasks: rather few things outside video games are).
And reducing startup time matters if you care about n-9s uptime.

(Of course, if you cared about n-9s uptime you would be doing
something about Solaris patching, probably with the use of a club with
nails hammered into it.)

On Nov 19, 5:11 pm, ab...@127.0.0.1 wrote:
> On 2007-11-18, Rich Teer <rich.t...@rite-group.com> wrote:
>
> > On Sun, 18 Nov 2007, ab...@127.0.0.1 wrote:
>
> >> While I agree it's better than sysvinit in principle, in practice
> >> the only selling point is ability to restart dying daemons, which
>
> > I disagree. The ability to start multiple services in parallel is
> > useful too,
>
> Well, maybe on a "can run 48 threads at once but only at 200MHz"
> Sun box. How long do your services take to startup so that parallel
> vs serial makes noticeable difference?

I use to compare boot times on my dualboot laptop.

In Solaris I run everything that is default installed + two zones, two
database servers, two webservers, and one Sun application server.

In WinXP-Pro I run default stuff + java.

I compare booting, logging in and start Thunderbird and I stop the
clock when connected to my mailprovider.

Guess who is fastest to that point.

On Mon, 19 Nov 2007 10:42:16 +0100 Miroslav Zubcic <news@big-other.com> wrote:
> Frank Cusack wrote:
>
>> But really I was trying to question the earlier poster's judgement.
>> svc > SYSV init > BSD rc,
>
> But you failed to question it. Where are the tehnical arguments?

I did question it. You're right though, I did not argue my side.
Let me review some of the other posts before responding.

....
>> ps. I used 50 b/c a friend of mine who i've been trying to convince
>> that IM and SMS are useful tools recently told me sorry he can't learn
>> to operate/think that way,
>
> He is right in SMS case. SMS is PITA to type on mob without proper user
> friendly interface like keyboard.

Kids don't seem to think so. I've always had a keyboard (e.g. now I
have a Nokia e70) so haven't hit that problem. But that's "just" a
device problem, SMS is a great tool regardless. My friend's complaint
was not about his phone. He is the type of guy that uses voice tech
support for detailed technical problems (the type where email of
config files, etc. is far far better than trying to tell someone over
the phone what is wrong).

On Mon, 19 Nov 2007 09:23:29 +0100 Miroslav Zubcic <news@big-other.com> wrote:
> Frank Cusack wrote:
>
>> On Fri, 16 Nov 2007 01:05:28 -0600 Chris Cox <notccox@notairmail.net> wrote:
>
>>> The fact that some many enterprises are running RHELAS 3
>>> sort of proves that it is stable... whether we want
>>> to believe it or not.
>
>> Don't make me laugh.
>
> Of course, you have some facts and arguments behind this laughing?

Just my own experience. Not with RHEL3 specifically (earliest I've
used is RHEL4), but in general I find Linux to be unreliable. I have
extensive experience at what may be the largest Linux deployment in
the world (and would have been at the time I was there also).

While I don't agree that RHEL3 is stable, my argument was NOT whether
or not it is. It was that whether or not people run it proves
nothing.

-frank