> 
> On Sat, 23 Sep 2000 [EMAIL PROTECTED] wrote:
> 
> >My main point is that this is not proprietary Unix where cost
> >virtually ensures only relatively big organizations (the needs of
> >smalleer ones could be handled by cheaper systems) use it and thus you
> >can design basing on the paradigm the system adminstrator is a
> >dedicated one.
> >
> >Since Linux is cheaper it can go where Unix can't go and constraints
> >in smaller organizations are different.  
> >
> >About the hiring of a consultant:
> >
> >1) What if the nearest one is at 60 miles in bad roads?
> >2) What if the user's job could require intervention at 9 pm?  This is 
> >charged extras.
> >3) Why the user should pay for what is objectively a design blunder (ie
> >designing without paying attention to his constraints)?
> >4) Which car would you buy?  The one where replacing a punctured tire or
> >filling the gas tank requires you pay a trained garagist with special tools
> >or the one you can do these things yourself?
> 
> You just don't get it.  There is _NO_ 100% solution that will
> stop every single user from ever experiencing disk corruption,
> and hence fsck from needing to be ran.  It is impossible for fsck
> to automatically repair filesystems 100% of the time in an
> acceptable way without manual intervention.  Therefore, if
> someone can't use fsck, doesn't want to reinstall, and can't pay
> someone to fix it for them, then they shouldn't use a computer.
> 

The fact there is no 100% solution is no excuse to not take steps to
reduce probability of problems.  One way is to improve the init
scripts, the other one is to shift to a journaled FS (not necessarily
Reiser) ASAP.

And as I said you can't design for Linux where you have individual
users and small organizations like in Unix were such are virtually non
existent.

Let's see a concrete example: fonts.  RedHat uses a font server.
Problem: what happens if font server has been stopped or does not
start due to a corrupted font path: X does not start and in case box
is at runlevel 5 there is an "X loop of death" where init indefintely
tries to start it.  Unless you are in a LAN and telnet is started you
could end resetting the box (means an fsck) since it is very difficult
to CTRL-ALT-DEL this box just the right moment.

If you live in a Unix world where the system administrator is ever a
user with years of experience this is no big probelm since you know
that a) he will not make stupid things like stopping the font server
and b) he will able to cope with problems

if you live in a Linux world where people could be individual users
caring for themselves from the start then you know some of them will
stop font server thinking they don't need it and that they will unable
to handle the situation.  In that case you start X with a fixed font
path and in its setup script try to replace it with a font server (if
font server is not running this will fail).  This is a _one_ line
patch.  This way even in case font server does not work, X will not
crash.  Since no font server means no True Type fonts you could
improve this by having a window popping up if font server is not
running.  Trivial.

Does the above solve all X problems?  No, but it makes X more robust
and robustness is a must given Linux sociology.

-- 
                        Jean Francois Martinez

Project Independence: Linux for the Masses
http://www.independence.seul.org



_______________________________________________
Redhat-devel-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-devel-list

Reply via email to