[svlug] This mess is surely a conspiracy, my brain hurts and I don't like complexity or live long and prosper

Sanatan Rai sanatan at gmail.com
Thu Jan 1 15:22:30 PST 2015


On 1 January 2015 at 17:43, Ivan Sergio Borgonovo <mail at webthatworks.it>
wrote:

> Complexity is unavoidable, I'd say invaluable actually.
>
>
This isn't about complexity. This is about changing things but not
necessarily making them better. SysV init is well understood. People know
where to look when things go wrong. I don't claim to be an expert on
systemd, but using timestamps that need to be converted to something
human readable in dmesg---now that it seems to be the default log---is
perverse.

One of my machines sporadically drops the network connexion. This is
easily fixed by a restart of network services. I see no errors in syslog or
dmesg. I'm sure that I can find the problem once I figure out the
appropriate
log to look into. This used to be easy. Now it isn't. I will eventually
find the
time to figure this out, but my point is that I don't see this approach as
making things better (for me or non-sysadmin users) in any noticeable
fashion.

One could go into nauseating off topic detail about why a system bus is
a good or a bad thing, or indeed why an object broking architecture is a
good or a bad thing. There are use cases when each of these things works
well and use cases when they're really bad ideas. I don't see how DBus
makes my user experience better. It isn't fast. I'd much rather not have an
extra process whose sole job is to make X desktops/apps easier to write.
If someone wants to use DBus, good for him. I should be able to run my
boxes and have a decent window manager without needing DBus. And no,
I don't regard tiling window managers as progress. The fan boys will
tell you how fast they are, and how clean. And all because they're coded
from scratch in C/Haskell---and so they don't depend on cruft like DBus,
gconf  and their ilk.

There are some investments that requires you to make debts and once
> you're mostly successful you may even forget you've some small debt
> hidden in some corner. sysvinit is actually a reasonably old system debt
> for Linux and systemd is not the first attempt to improve it.
> Worse is better, better is worse and everybody agree we can disagree,
> but someone at the end has to be right, so really worse is better is not
> really true, it is actually false and surely now someone is wrong on the
> internet.
>

I'm not objecting to change at all. I'm just saying that change that doesn't
necessarily lead to clear and reasonably immediate improvements is a
waste of time. The first tractors had reins instead of a steering wheel
because designers thought that they'd be familiar. However, they sucked.
Soon the steering wheel became standard and I think we'd all agree that
it's superior for piloting motor vehicles (though not horses). So the
irritation
is not with the change, but with a change that seems to make many
commonplace tasks difficult, without adding a noticeable benefit.

Did anyone came out with the perfect method of software development in
> an environment that is as rich as complex? [2]
>
> Do we have the perfect software? No
> Do we have a better software than say 5 years ago? It seems so.
> Do we have to be concerned? It just depend on how serious you are about
> this particular problem, because there are many others.
> If you've the skills do something, if you don't have the skills get
> them, if neither invest your time better.
>
> As a side note unix command line utilities don't really shine for a
> coherent powerful interface, that was secretly meant to make them really
> decoupled with anything with any random interface, IO redirection was
> there just to mislead you.
>

Perhaps. However, do you have an alternative? Eg, PowerShell's approach is
different, in the sense that it purports to be consistent. Unfortunately,
they decided on the `everything is an object' model instead of 'everything
is a (text) line', and that's a lot more annoying. So

wc -l CL*.log

becomes

get-content CL*.log | measure-object -line

[I'm actually not sure if the above just works, but it shouldn't be far
off.]
It's very consistent, and you can use all the `power of .NET', but it sucks.

 |Working with JS is much nicer now than it was 10 years ago, and you've a

That's a bit like saying that you can get used to banging your head against
the wall and even like it. ;-)

lot of tools. Still JS compared to python is a PITA and people are
> writing "compilers" as CoffeScript to avoid to directly deal with JS.
> But JS is in all browsers and in all mobile devices (and Dart isn't yet)
> and it will take time to replace it with something better... or with a
> better JS.
> I think very few people are writing Java applets that run in the
> browser, Flash is unfortunately still alive but Adobe and Apple sent a
> signal it may disappear (like COBOL), javascript is still among us and
> strong. The irony.
>

Ie  bad choices made now will be with us 10 years from on. All the more
reason to choose carefully.


>
> Should we talk about how the story on client and server went? or the one
> about process/threads/virtualization/containers?
> What about RISC/CISC/CPU/GPU?


Go on, I'd be interested!

-- 
Sanatan Rai
3, Admirals Court,
30, Horselydown Lane,
London, SE1 2LJ.
+44-20-7403-2479.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.svlug.org/archives/svlug/attachments/20150101/4b037508/attachment.htm


More information about the svlug mailing list