[KLUG Advocacy] Re: Nautilus in RH 7.3 and Samba shares...revisited... -- years of NT/UNIX corporate network experience ...

[Bryan J. Smith]

--=-h7o2TdqwG2Ae7tgEpDoK
Content-Type: text/plain
Content-Transfer-Encoding: quoted-printable

On Sat, 2002-12-07 at 16:05, Adam Williams wrote:
> I tooks his comments about IE, etc... to be in reference to the
> architecture (modularity, component-ness, etc...) and not that it was an
> ideal implementation of those ideas.

And I do too.  But the underlying "this components depends on this other
component" is a problem when you're talking critical, system-level
software.  Yes, for end-user apps, go for it!  But _not_ for network
services.

> So many things (GUI, etc...) are tied into the kernel in Win32 it
> seems obvious that they would have trouble doing a truly component
> based AND secure implementation.  Linux (IMHO) has an advantage there
> that many things (GUI, etc....) are already distanced from the
> kernel.  And from the applications I've used Bonobo is a great
> success.  Much of the same code that is Bonobo is going into Mono.=20
> GNOME's equivalent to ODBC for generic database access, as well as
> data-awate widgets, GNOME-DB; is the database provider for the Mono
> implementation.

But I still have some reservations on making some things more
"independently piecemeal."  Anyone who has run a Gnome desktop on an
NFS-mounted home directory knows what I'm talking about.  One failure
with Bonobo and you're screwed.  Yes, I know they've addresses things
with Gnome 2.0, but it's still a mentality that leads to a cascade of
issues.

Miguel is basically implementing a "registry" with GConf.

> I think a few idealists around is a good thing,  they come up with great
> concepts,  which are realizable to about 90%, and everyone benefits.=20
> I'm a pessimist and quite cynical,  thus the universe has balance.

Well, I've been around Microsoft a long time.  Now Miguel doesn't have
the same, "this should be done, but management says no," problem that
Microsoft developers run into, but the .NET model still leaves some
questions.

E.g, the "common language run-time" concept that is supposedly
"superior" to "single langugage Java" is _flawed_ because it _assumes_
you can freely use code between them.  Not!  You're non-C# code is very,
very limited, C#-centric, and basically can't be object oriented (e.g.,
no inheritance).

So all we have is a Win32-centric version of Java in .NET.  Yeah, there
are a few improvements to the Java language, but that's just hindsight
at work.  Nothing that couldn't be implemented in a .jar file as someone
discovered.

In fact, what really disgusts me how Miguel talks about the .NET
security model being "better" than Win32.  Win32 was actually
_very_good_!  The "problem" was that _no_ developers adopted it because
it was _only_ supported on NT-based Windows platforms.  So if you wrote
for it, the application wouldn't run on DOS-based Windows (95, 98, ME),
and most developers (even Microsoft's own application division!) didn't
bother to support it.  [ I think the _only_ company that did was Digital
-- who basically wrote most of NT and _all_ of 64-bit NT ]

But I do agree that Mono _is_ important from a portability standpoint.=20
As more and more Windows developers move to .NET, we'll need an
equivalent to "WINELIB" (_not_ WINE, but the porting libraries).  Mono
will provide that.

> Sure, thats the point of CORBA, Bonobo, GNOME, Mono, etc...  Emphasis on
> flexibility;  as the main problem with Linux/X/GNOME applications is
> still that they don't play together without real work.  GNOME
> applications less so then straight X/Linux apps,  but thats because many
> Open Source developers don't "get it";  they can't blame the technology
> any more.

But even now Miguel has "switched" from focusing on CORBA to .NET.  It's
almost like the GNOME team is "trying out" different things to see what
works.  All it's doing is following the 100% Microsoft "problem" of,
this technology today, that technology tomorrow, oh-crap, we've now
gotta support both of them in yet a third version.

Now don't peg me wrong here.  I _love_ Gnome from an user-interface
standpoint, and canNOT stand KDE.  But the KDE guys took the mentality
"let's just get the apps out," came up with KParts, and the damn thing
works pretty good.  And it's less "interdependent" on other components
too.

Again, that's the problem of an idealist.

> You can, but your going to get it one way or the other.  What is gconf?=20
> A registry.

Yes, I know.  I don't agree with the approach either.

> I don't like M$'s implementation of their registry,  but in
> the end every workable system has to have one.

Give me a directory tree with an index.  That's the most reliable.

I'm currently working on some archiving/backup software.  In the
archiving/backup work, there are two mentalities:

  Streaming:  cpio/tar/pax -- meta-data is stored sequentially with data
   Indexing:  zip/others -- meta-data is a single directory after data

The first is most recoverable in the case of even multiple errors, but
slow in random access.  The second allows very fast random access, but
can leave an archive unrecoverable in the case of a single error
(creating a "single point of failure").

Why oh why haven't people tried to use a combination of both?  [
Actually, that's what a new format I've created does -- long story ]

It's like the chronic problem with "all-in-one management tools."  While
they give you all your tools in one place, they don't always do
everything that the native tools do on their own.  But if want to use
the native management tool for a system/service, you don't always know
what it is.

Usually the "key" is to just build a basic index/listing of all the
system/services, and then launch/access the native management tool for
them.

> Applications to be truly useful must inter-operate,

Not always.  The concept of "inter-operate" is often a mis-understood
one.  Sometimes "inter-operable standards" are "overblown" with a basic
text file, or messaging passing protocol would do.

> components must be located so they can be used, etc...  User's
> preferences need to be published in some global way.  I like
> GNOME's move to an XML registry, almost as a volatile database.

And they seem to take the "single point of failure" path that Microsoft
often takes.  That's bad.

> And in another sense LDAP is very much a "registry", just a
> network one.  Now if gconf could store all those settings in the user's
> LDAP object we would have the ultimate in "roaming profiles".

The idea of "roaming profiles" is a management headache.  Yes, GNOME
will do a better job of it than Microsoft (who's different departments
have "disagreed" on their own implementation over the years, resulting
in different, incompatible approaches), but it will still have many of
the same problems as Microsoft Windows.

> My profile could follow me to any GNOME desktop IN THE WORLD with
> the help of a few SRV records.  That would be fantastic.

Again, nice idea.  But the Miguel-GNOME approach follows with Microsoft
ideals a little too closely.  ;-p

[ Remember, I was an original NT 3.1 beta tester, and have seen this all
happen before. ]

> How many uses are using the typical desktop/laptop/workstation?

The problem is, there is _no_difference_ between a "workstation" and a
"server" from a security and management standpoint.  That's why
Microsoft technologies _suck_, because they don't realize this!

> I agree, except that M$ is easy-to-use.  I can easily more time farting
> about with stupid M$ problems, then it would take me to build a
> corporate E-mail infrastructure.

I can easily point-n-click my way to obliteration.

As I always say, "I think it's great that anyone can modify Windows
settings very easily.  But how do you account for changes?=20
Point-n-click still does not solve the greater problem with chanage
management.  In fact, it often compounds it."

A concept as simple but extremely effective as "ci -l" (RCS check-in) on
a configuration file after modification is extremely foreign to the
great majority of system administrators.

Just based on my years of experience managing NT/UNIX corporate
networks.

--=20
Bryan J. Smith, E.I. (BSECE)       Contact Info:  http://thebs.org
[ http://thebs.org/files/resume/BryanJonSmith_certifications.pdf ]
------------------------------------------------------------------
  The more government chooses for you, the less freedom you have.


--=-h7o2TdqwG2Ae7tgEpDoK
Content-Type: application/pgp-signature; name=signature.asc
Content-Description: This is a digitally signed message part

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQA98mjfDjEszaVrzmQRAtieAJ96qK841gkasX16KBGtNNOT4iO2gQCfY/W5
fMlyBBk4ePtxIq8elYkh3Fw=
=U8sZ
-----END PGP SIGNATURE-----

--=-h7o2TdqwG2Ae7tgEpDoK--