On 04/02/2015 09:27 PM, Lennart Sorensen wrote:

On Thu, Apr 02, 2015 at 04:15:10PM -0400, Alex Volkov wrote:

...

I thought it would be relevant to this list. Among the new features --

https://fosdem.org/2015/schedule/event/whats_new_in_systemd,_2015_edition/

Things recently added and being worked on:

* systemd-nspawn -- docker-like containers * btrfs support * firewall support based on application id and not port numbers * moving kernel console into userspace, adding high DPI and proper Unicode support * resolving config file path with sysctl --edit / sysctl --cat * network management with networkd * auditing has indexing and is included in logging * support of stateless systems * resolvd -- vpn support, smart dns with support per-interface cache, support for LLMNR * rolling avahi into systemd is being worked on * journald remoting -- remote logging with http using pull and push model * logging stack traces without gdb * protect system/home/tmp/devices -- setup minimal namespace environments for services * timesyncd -- trivial sntp client with time saving for the devices with no rtc * autodiscovery of swap partitions with gptd * readahead support was dropped * merge gummiboot into systemd (UEFI secure boot support) I can see why people complain about featurecreep.

If you're building primitive functions as the main effort, and shipping products sitting on top of thgem, this scales. Even elegant constructs like qef have a ton of stuff of this sort. If, however, you're writing A dpends on B depends on C depends on A for any kind of A, B and C, you're creating spagetti code ( or perhaps a walled garden if it is intentional). Both lead to business failure. --dave -- David Collier-Brown, | Always do right. This will gratify System Programmer and Author | some people and astonish the rest davecb@spamcop.net | -- Mark Twain

On Fri, Apr 03, 2015 at 10:13:23PM -0400, Christopher Browne wrote:

I'm glad that it is a sufficiently vital project to have things going on.

I'm pretty happy to see them adding things that smell like extra services, though this shouldn't require "kernel of systemd" involvement, surely. For instance, if timesyncd is a wee service where they combine some shell scripts (or such) with a service configuration file, that's great.

I'm less enthralled with the vast variety; it does indeed look like "scope creep."

Explain to me WHY it needs BTRFS support, why this isn't better separated from systemd concerns. I gather that it's intended to help support the "containering" support, which might be OK, but when scope creeps, I get a bit scared that the service is trying to be the "creep" of Linux, and that's no good thing.

I suspect that just like systemd has container and virtual machine management support, it is now adding btrfs volume and snapshot management support (some of which might even be useful in conjunction with containers or virtual machines). It does seem like trying to tie too many things together. Had you noticed there is a machined virtual machine manager in systemd? I only noticed that recently. And of course machinectl to go with it. You can even do systemctl -M vmname command, to run systemctl command inside that named vm. I suppose that could be convinient. -- Len Sorensen