Re: [tlug] State of "Linux" documentation [C&C]

[Joe Larabell <fred62@example.com>]

On Tue, 10 Jun 2008, Stephen J. Turnbull wrote:

> ... My point is
> that *unlike cars* almost all of what is under the paint (including
> the metal of the hood!) of one GNU/Linux is the same as all the
> others. So when "skinning" the system the distro vendors have done
> their users a disservice by making it ever harder for them to see how
> that is being done, or even for a geek buddy who has memorized the
> manpage for ifconfig(8) on 6 different systems to figure out how eth0
> gets its IP address.

When I installed my first Gentoo system many years ago, I was forced to 
make decisions on things which I never realized were even open to any 
decision. Who would have thought that there were multiple ways to write 
the system log -- or multiple versions of cron? Not just "versions" in 
name only -- very different implementations with different config files.

I picked syslog-ng and vixiecron. I'm satisfied with my choices but I'm 
not convinced any of the others would not have been just as satisfying. In 
fact, I'd be willing to bet someone out there is reading this and saying 
to themselves (if not to the list, eventually): "No... <whatever>-cron is 
far more advanced than vixiecron". And, given the number of years that 
have passed since then, they're probably right. But I know these tools
well by now, they work for me, I know the config format. Why change?

My point, I guess, is that there isn't quite as much in common under the 
hood as one might suspect. There is no "standard" for how ipconfig should 
work so when someone gets a new idea to make it work better, they write a 
new version. If everyone *else* thinks its better, the old one falls into 
obscurity. That leaves me free to pick whatever variation I like from the 
choices available at the time. And, presumably, someone does exactly that 
for each of the major distros. If all the UIs and config files were cast 
in stone, there would be no room for innovation.

The downside is that it's no longer possible (if it ever was) to document 
a complete "common" Linux system because everything but the kernel can be 
swapped out whenever something better comes along (in all fairness, the 
kernel can be swapped out too -- ie: BSD, etc -- but doing so takes the 
system out of the realm of the current "Linux" discussion ;-). About the 
only things for which there are *not* multiple implementations (AFAIK) are 
the compiler toolchain and the basic shell utilities like "cp" and "mv". 
And I suspect that the only reason those are common among all distros is 
because nobody has gotten annoyed enough at how they work to re-implement 
them to work any differently. But, if someone did, there goes that common 
Linux documentation out the window.

(New users, by the way, don't generally *want* to know about "cp" and 
"mv". Whether that's a good thing or not could certainly be debated.)

> ... GNU/Linux could provide a lot of introspective
> capability into the common implementation, and the distros could
> provide UIs that organize it and make reliable and accurate access to
> most important aspects.

I'm not so sure that's something to lay on the distros, specifically. 
Anyone could implement a system config UI and release it as open source. 
In fact, I believe there are a few already. That they are not in common 
use across the board probably speaks more to their lack of popularity or 
maybe to users preferring the status quo than to any failure on the part 
of the distro maintainers. Why isn't Webmin, for example, used as the 
primary config UI for all distros? Would that be preferable? I don't think 
so, but... I'm one of those who can find my way around without the book so 
I'm not likely to use the UI anyway.

On Tue, 10 Jun 2008, Edward Middleton wrote:

> The point is that basically all computers are the same, at both a hardware 
> and software level.  If you look at Unix's they are all extremely similar.

Similar, sure. Alike, no -- not enough alike to not confuse beginners. Not 
when you consider the whole constellation of utilities that can easily be 
layered on top of the common kernel (which a beginner shouldn't have to 
deal with anyway). And if we're not talking beginners, then there's tons 
of books on various bits and pieces of the Linux infrastructure. Too much, 
in fact, to ever stuff into a single volume. So you have distro-specific 
hand-holding books to get you started and, once you start to grok the bits 
and pieces, you scan the O'Reilly catalog for in-depth material on the 
specific "common" pieces in which you're interested. I'm not sure how we 
could do any better.

> You can obviously learn about a particular Unix by focusing on the unique 
> implementation of some features, which is what you will get in a distribution 
> specific book.  If you do this, every new Unix you learn will be a wonderful 
> mystery ride, but who has the time for this?

When you've just unpacked your system and you're trying to get the printer 
to work, who has the time to read the entire sordid history of printing on 
Linux? For a beginner, it would be a disservice (IMHO) to give him much 
more than some screenshots, step-by-step instructions for the distro he 
has in front of him, and maybe an appendix called "Once you're ready to 
swim in the adult pool, read these...".

Also, one could argue that if you end up buying more than one of these 
distro-specific books, you're probably not the type who's likely to learn 
the common stuff anyway ;-).

IMHO, as always...

---
Joseph L (Joe) Larabell            Never fight with a dragon
http://larabell.org/                    for thou art crunchy
http://thelemicleague.org/        and goest well with cheese.