| 1 |
On 22.02.2014 11:40, Mark David Dumlao wrote: |
| 2 |
> [ ... ] |
| 3 |
> Even as the complex beast it has become systemd is still simpler than the |
| 4 |
> alternative of having abominations of unreliable shell scripts checking to see |
| 5 |
> which version of grep and sed is used to split the command line, or whether |
| 6 |
> the system uses tempfile or mktemp, or depending on perl. |
| 7 |
|
| 8 |
Well, simpler yeah, supporting only one kernel of specific versions is |
| 9 |
always simpler than trying to support everything from SunOS to NetBSD. |
| 10 |
This way, if the kernel supported only e.g. Intel IvyBridge+ with one |
| 11 |
chipset family, one graphics (VESA) and so on, it would have been |
| 12 |
incredibly simpler. |
| 13 |
Or probably is it systemd that checks "whether the system uses tempfile |
| 14 |
or mktemp"? Or having a new own (NIH) unit config format parser is |
| 15 |
simpler than taking one of thousands of existing ones? |
| 16 |
It is simpler for you end user. (Though dubious for users who wield |
| 17 |
shell scripts and perl.) But when it comes to reliability it's entirely |
| 18 |
wrong. I can fix a faulty shell script without having to wait for a new |
| 19 |
release. Or I can even write mine own. Can you fix systemd? |
| 20 |
|
| 21 |
> ergo libreoffice. |
| 22 |
|
| 23 |
A follower of the MS-maintained strategy "one black box that does |
| 24 |
everything". I always wonder what does e.g. Excel/Calc/spreadsheed need |
| 25 |
font decorations for? Or 1+2 is different from *1*+/2/? ;) |
| 26 |
|
| 27 |
> desktop environments. |
| 28 |
|
| 29 |
A good DE design is just a collection of separate tools doing different |
| 30 |
things, united by some graphical design. |
| 31 |
|
| 32 |
> firefox. |
| 33 |
|
| 34 |
An example of how an app once followed a non-Unix way is now failing to |
| 35 |
get back to the Unix way. And its reliability is somewhere near zero. |
| 36 |
Though, it's almost impossible now to make a simple browser which would |
| 37 |
also be cross-platform and popular. There's a holy bible of standards |
| 38 |
and quirks to support, and yet more in the development phase, for the |
| 39 |
end user to be happy. It's completely different from an init system, |
| 40 |
even all the init systems of all Unixes altogether. |
| 41 |
|
| 42 |
> databases. |
| 43 |
|
| 44 |
What's wrong with them? Mostly, characteristic examples of the Unix way. |
| 45 |
|
| 46 |
> Heck much of |
| 47 |
> what's being said about systemd applies to postfix - there's no general case |
| 48 |
> reason for me to grab some random postfix component and use it for everyday |
| 49 |
> work, therefore postfix is just some closed-source monolithic virus, right? |
| 50 |
|
| 51 |
So now that there's a working postfix which is an example of a non-Unix |
| 52 |
way design, it's justified to use this approach everywhere else? |
| 53 |
|
| 54 |
|
| 55 |
-- |
| 56 |
Best wishes, |
| 57 |
Yuri K. Shatroff |
Archives