:: Re: [Dng] John Goerzen asks, "Has m…
Página superior
Eliminar este mensaje
Responder a este mensaje
Autor: Steve Litt of Troubleshooters.Com
Fecha:  
A: dng
Asunto: Re: [Dng] John Goerzen asks, "Has modern Linux lost its way?"
On Wed, 11 Feb 2015 23:17:36 -0600
"T.J. Duchene" <t.j.duchene@???> wrote:

> "John clearly states that he believes the problems are distinct from
> systemd. While many here may not necessarily agree, I do agree that
> various aspects of the system have become, if not complex, at least
> more opaque than in the past.”


I totally agree. I think systemd is just the worst of many entangled
monolithic monsters, and is a symptom of the true problem.

The preceding sentence does not in any way lessen my resolve to fight
systemd any way I can.

>
>
>
> You're right. I think the problems, and frankly systemd as well, stem
> from the fact that the community has changed. I've been working with
> computers and writing code for a very long time, over 25 years, so I
> think that I am somewhat qualified to make that statement.


Agreed.

>
> There was a time when the “lingua franca” of the community was C. If
> you were going to be part of the community, it was expected that you
> would become proficient with it. Since everyone understood C, no part
> of Unix was opaque, from the kernel to userspace. Things have
> changed. Now, people use anything from Python to C#, with a minority
> using C regularly. Most are kernel developers. Suddenly, everything
> not written in the favorite language of the day becomes opaque.


I agree that use of C has lessened, and that less of us understand C
today. I disagree that this is necessarily the cause of the problem.
Personally, I think that anything that *can* be written in Python,
Perl, Ruby or Lua *should* be written in one of those languages,
because doing so limits buffer overruns to those remaining in the
language. Also, software written in those, especially Python, which has
a rich library of standard libraries, has fewer dependencies that need
to be installed. And yes, feature for feature, they're more readable
than C.

Also, the root cause of the gross overcomplication isn't that many can't
read and write C, it's that many are clueless of the operation of a
computer beyond recipes of mouse clicks. Naturally, such clueless users
don't know C, but the real problem is they don't know directories,
don't know pipes, don't know elementary Unix filters, etc.

>
> I'm not trying to start a flame war by saying this, but I think at
> least 1/3rd of the problems people have with systemd is the fact it
> is written in C.


I'd blame it more on this:

http://en.wikipedia.org/wiki/Systemd#mediaviewer/File:Systemd_components.svg

It's the architecture, plain and simple. It's a V'jer like
conglomeration of clinging junk. Worse yet, it subtly changes simple
softwares that you could formerly just plug into the system like
electronic ICs. With systemd, in for a penny, in for a pound. Plus
take a look at the plans Lennart has for you:

http://0pointer.net/blog/revisiting-how-we-put-together-linux-systems.html

If you like DIY Linux, you hate systemd, because systemd has removed
interchangeable parts, replacing them with parts exactly machined for
the systemd system, and there are no other such parts. Well, there
wouldn't have been if it weren't for the Devuan project, and I thank
you all for that.

> If it was written in Python, I have my doubts that
> it would have created such a stink. There are large, complex pieces
> of system software written in Python scattered all over the Linux
> community and no one treats that as an apocalypse, even when they
> cause huge stability problems.


I personally know of no Python *system* software, but even if there
were, I'll bet you such Python programs are single units that can be
easily replaced by a good programmer. Such a replacement can implement
the functionality in a different language, or even add specific needed
extra functionalities.

Contrast this with systemd, which requires knowledge of all sorts of
other things, and Lennart's plan is to have systemd incorporate more
and more of our computers' functionality, leading to a situation where
1 hour quickie programs are a thing of the past.

> “I think what has bothered me the most over the past few years is the
> churn and what sometimes seems to be adoption and then replacement of
> a technology without explanation”


I'd be more succinct about the churn. Most of the churn I see is
pandering to the point and click crowd, trying to make Linux ever more
like Windows and Mac, so that the brainless can use it. As opposed to
what we used to do: Offer a sane and discoverable system that brought
new users to a new understanding of computers.

Bill Gates and Steve Jobs profited from convincing people that they're
too stupid to use a straightforward interface. We have no such profit
motive, and therefore should not pander to the stupidest of new Linux
tire-kickers.

>
> This is symptomatic of the way Linux is developed in distributions.
> Rather than agreeing to common standards, each goes their own way.
> This is not a bad thing – not at all – but it tends to be disruptive
> from time to time.


Lucky thing for me that they differ. I don't have the time to
compile/install *too, and I don't have the attention span and
attention to detail to do Arch's 40 step manual install recipe. Nor do
I have the patience to deal with Slackware's "we don't need no steenkin
dependencies" package manager. And yet, I have friends who wouldn't be
caught dead using any of the distros I use (Manjaro, Wheezy, Ubuntu,
and hopefully soon Devuan). One man's opinion: The distro differences
present us with much needed choice.

And really, choice is the key to the entire discussion. Nobody on this
list is saying that systemd should be banned from this earth. We're
just saying that *we* don't want it on our computers, and we're working
toward that goal.

SteveT