Public SourceMage Discussion List

["Sergey A. Lipnevich" <sergey AT optimaltec.com>]

All,

It's been a frequent argument in my conversations with other developers that
we're about choice. That is true, but only if we know where to stop. Us being
pro-choice doesn't mean we get to build infinite customizability into every
nook and cranny of the distribution. We've got to know better.

Take security. We're about choice, right? Why don't we leave it for the
discretion of the user whether they'd like to patch an insecure piece of
software or not? Why don't we ask in CONFIGURE about every security patch,
would you like to install it? The reason is common sense: we know and care 
more
about where the choice stops. The users, knowledgeable or not, don't need to
spend their brain cycles on this question because we are qualified to make 
this
decision for them. This reasoning does not inhibit choice.

Choice can lie dormant in a spell, but it's still going to be confusing. For
example, I created http-dev because I needed Apache 2.1 with new 
authentication
system. Now, it shows up when people need to install a Web server. It's
confusing to see apache, apache2, and something else that kinda sorta is the
same thing, in the list, isn't it? Choice like that is an example of developer
or sysadmin (me) looking for easy way out. On the surface, it creates more
choice. Below surface, it creates even more confusion.

Another example. Casting evolution2 creates an avalanche of choices for its
dependencies, and if you agree to more of them, you'll get another avalanche
from gst-player, and another from qt-x11, and so on. How easy is it to get
confused? I'd say it's almost impossible not to be. In this case, choice is 
the
result of our distro's design, and can't be easily simplified. We fight it 
with
default answers (that do not always make sense, BTW).

Next time we discuss what change to make, I suggest we try going to the forums
or IRC and validating the claim that users need the stuff you suggest.

At this point, I see two trends. We drift away from being a sysadmin-friendly
distribution, we drift away slower from developer-friendly, and we become
desktop-oriented with majority of spells added for this area. The second, more
arguable trend, is that we increase complexity without adding more value for
users. I basically operate my servers the same way since 2001, no new features
were added since then that would allow me to be a better sysadmin. As a
SourceMage section maintainer, my way of doing it also didn't improve
significantly (I still create new spell by coping existing ones, I still do
every version upgrade by editing not one but two files in three different
places, and thank you for md5unpack function 'cause that the only true
improvement).

I may be wrong on both of these trends, these are very subjective, but this is
how I see it from my place.

So, in our pursuit of choice let's not forget other things, like quality,
stability, predictability, easy of use, automation of mundane tasks, etc.

Sergey.