[olug] linux web server management ?

[William E. Kempf]

Jacobs, Robert A. said:
>
>
>>-----Original Message-----
>>From: William E. Kempf [mailto:wekempf at cox.net]
>
>>  In other words, from a developer's stand point the Win32
>>Registry API is a god send,
>
> ...and from a user's point of view it is HELL (disclaimer: I'm not a
> Win32 Developer).
> The Windows Registry is not a good idea -- it is the most obvious single
>  point of failure on a Windows system.  Corrupt the Registry and you're
> done.
>
> Eggs? Basket? Anyone?

Yes, but why do you think I have repeatedly said that the "monolithic
single binary" is a bad idea.  The registry _concept_ is good.  It's the
registry _implementation_ that's flawed.

> The Registry is more arcane than any configuration approach or file I
> have yet to see in Linux.  Made a change? Want to go back to your
> previous setting?
> Didn't export your registry before you changed it?  Whoops.  Sorry.

Uhm... that's not really any different than forgetting to make a backup of
a configuration file before editing it.

> The Registry may be great for developers but it is practically useless
> to users....

I disagree entirely.  Registry hackers make use of the registry very
frequently.  And since the API is standardized, they've been able to
*easily* create GUIs, such as TweakUI, for those who are scared to touch
the registry directly.

But Linux can take this idea and make it better.  Don't make it binary. 
Don't put all the configurations into a single _file_, as a single point
of failure.

> and sometimes, hacking the Registry is the only way to make certain
> configuration
> changes, optimizations, etc.

Rarely, and when this is the case it's because the programmer didn't
intend for you to use the configuration.  The same thing can be said in
many applications using the "Linux" approach, where they don't document
the settings.

> One of the greatest things about the way Linux approaches application
> configuration
> is that each tool knows where to find its files.

And this differs from Windows how?

>  When I install
> application foo into
> /usr/local/foo and later delete the foo directory, foo and all of its
> files go with it.

Uh huh.  I've got a lot of config files scattered around my Linux box from
uninstalled packages to prove you wrong on this one.

> The only thing that *may* remain are some dependencies I installed while
> installing foo.
> I don't run the risk of corrupting my registry or leaving dead leaves
> behind in it which
> allows me/you to keep a lean, clean machine.  The leaner the machine,
> the less apt
> someone is to compromise it.

I haven't argued any of this.

> Separation between applications is a good thing.  It is the "unity",
> interdependency and
> cross-application awareness of Windows applications that contributes to
> the many problems
> Windows systems experience and contributes to their reputation for
> instability and lack
> of security.

*shakes head*  I don't think you really understand the issues on the
Windows platform.  I can think of no situation in which "unit",
interdependency or cross-application awareness has contributed to the many
problems on a Windows system.  What causes those problems is *mostly* the
lack of versioning and a broken DLL system.  (And the single monolothic
binary registry does provide a single point of failure that can trash the
entire system, instead of a single app, though in all honesty I've never
had this happen in 10+ years of use.)

But I've *NOT* argued for Linux to do the exact same thing as Windows. 
Only that it should learn from the things that Windows has done right, or
at least better from an end-users stand point.

William E. Kempf