[sf-lug] Byfield's "Verdict" on systemd

Rick Moen rick at linuxmafia.com
Wed Feb 5 23:16:48 PST 2020


I wrote:

> They didn't listen.  [Joey Hess] quite soon after that.

As is probably apparent, that was supposed be 'quit'.


Quoting Alex Kleider (akleider at sonic.net):

> What do you (or others) think of UNIX and Linux System Administration 
> Handbook (5th Edition)? 
> Sobel's is another book that I've noticed is out with a newer edition 
> but someone who has a copy told me there's no mention of systemd in it.

I belatedly picked up on what you are _probably_ assuming, that the
obvious, right and proper way to learn system administration is to buy
and then sequentially read a good book on system administration.

Could work.  I wouldn't necessarily call that a bad idea.  I happen to
have not done it.

I moved laterally from technical support into what was then called MIS
(Management Information Services) in the 80s, and edged into the system
administration speciality mostly because freely distributable Unixes
(mostly FreeBSD and Linux) kept making my life simpler, so I learned
them.  I learned them mostly by using them and reading provided
documentation and guides, and by looking at example solutions published
to the Internet by others.

The classic books on system administration were the ones by Æleen Frisch
and the one by Evi Nemeth.  (Nemeth's is from Addison-Wesley, not
O'Reilly as I posted earlier.  Frisch is from O'Reilly.)   I got copies
but didn't get hardly any use out of them.  I also bought Cricket Liu's
classic book _DNS and BIND_, eventually realising that I was getting
hardly any use out of it.  For a while, I kept accumulating that class
of book, especially the O'Reilly Nutshell Series, until eventually I
spotted the pattern:  I was getting hardly any use out of them.  There's
still a whole lot of them on the shelves here at Chez Moen, decades
later.

Partly what happened is the Web.  Once you learn to do intelligent Web
searching, and how to spot and move on from badly written crud, you
_often_ find that what you can find from public sources is more than
sufficient or even superior.  (Of course, obsolete material lurks as a
hazard on the Web, but you _can_ get stuff that's very current.)

And the other thing that happened is that, as Linux became popular,
books about it started to suffer from what I call Que/Sybex Disease.

Remember back in the 1980s when the bookstores were flooded with 1,000+
page books, many of them published by Que or Sybex, that purported to
tell you all about, say, Lotus 1-2-3?  At the time, I realised that
there was a _huge_ contrast between the bugsquasher-length books for the
DOS/Windows proprietary-software market and leading books published at
the time about Unix:  

o  Pee-Cee app books were bloated to 1000+ pages.
o  Good Unix books, including early Linux-centric ones, ran about
   100-120 pages, maybe a bit more.

And I had to stop and ask myself why.  And the answer was obvious:  The
kitchen-sink aspirations of the Pee-Cee app book was an obvious
adaptation to proprietary software bootlegging.  Joe User made an
illicit copy of Lotus 1-2-3 or WordPerfect 4.2 from the office or a
friend -- and immediately felt insecure because he didn't have the
accompanying manuals.  So, Que and Sybex basicaly hired people to write
big-ass bugsquasher volumes for bootleggers to buy so they could feel
that, if they needed to, they could find anything they needed in the 
almost-a-manual book.

This also explained something that increasingly annoyed me about
Pee-Cee-market technical books:  The publisher and authors inevitably
declined to decide whether to write a tutorial or a reference, but
instead wrote a gargantuan single volume that tried to be both at once,
and filed.

In case the distinction is not obvious, technical documentation falls
into three broad categories:

1.  A tutorial is something you're intended to read front-to-back, 
sequentially, to learn the basics of a subject.

2.  A reference is _not_ intended to be read that way, but rather has 
in a semi-encyclopaedia fashion entries about pretty much anything you
might want to know about the subject.  The idea is that you _look things
up_ in a reference, not that you read it.

3. A quick-reference is a maximally terse reference, letting you look
up details of something super-quick, but with as little explanatory
detail as humanly possible.  A good 'man' page would be classic
quick-reference.  It's a bit masochistic to try to use quick-references
as if they were tutorials, but I suppose everyone tries.  ('I read the
man page about git, but I'm still not clear on how to develop with it.')

Anyway, cutting to the chase, any technical book that tries to be
simultaneously a tutorial and a reference is going to suck at both
roles, as inevitably those Que and Sybex books did.  _And_, as users
from the Pee-Cee market started getting interested in Linux, publishers 
started upping the page countes of each successive edition, and making
the books tragically fail by being both bad references and bad
tutorials.

5th edition of Nemeth:  1232 pages
4th edition of Nemeth:  1279 pages
3rd edition of Nemeth:   896 pages
2nd edition of Nemeth:   591 pages
(I don't have a figure for the 1st edition, but I'm not sure I've ever
seen a copy, either.)

Call me a cynic, but I'm pretty sure that half of the 5th edition's page
count, _at least_, is padding that is present to cater to the Pee-Cee
buyer's mindset and to provide a mental security blanket.  ('It's so
thick.  Everything I need to know must be in here somewhere.  I'm not
sure why I can't seem to find what I need.  It must be a problem with me
rather than with the book.')


So, I have a prejudice against bugsquasher books, because (1) usually
it's because of padding and the desire to be a refertutorial, and (2) 
all of that extra junk mostly tends IMO to make it slower and more
difficult to get useful information from.  And to that:  (3) Technical
books tend to go quickly out of date to some degree, especially if they
try desperately to be trendy.

-- 
Cheers,                     "Why doesn't anyone invite copyeditors to parties,
Rick Moen                   when we're such cool people out with whom to hang?"
rick at linuxmafia.com                        -- @laureneoneal (Lauren O'Neal)
McQ! (4x80)



More information about the sf-lug mailing list