3 ms·

Systemd is starting to show the signs of feature creep which IMHO is good. Failing early means the ship can be turned around if there's enough will, or at least

by erhardm 12y ago

Systemd is starting to show the signs of feature creep which IMHO is good. Failing early means the ship can be turned around if there's enough will, or at least (newer) replacements of systemd(in case it goes really off the rails) will learn from others' experience.

UnoriginalGuy 12y ago
This was added in 2012.
- erhardm 12y ago
  You mean the article or the feature? The article is dated by 22/11/2014. The feature may have been added a few months/years ago, but the point still stands about systemd showing signs of feature creep. Note also the DNS resolver issue[0] [0] - http://seclists.org/oss-sec/2014/q4/592 http://seclists.org/oss-sec/2014/q4/592
  - the_mitsuhiko 12y ago
    The feature was added in 2012 or something, whenever the log support came. The DNS issue affects not just systemd but any stub resolver.
    - pdkl95 12y ago
      The DNS issue is worth in systemd's resolver because it is NOT merely a "stub resolver" - it implements a cache. From the same thread: "The DNS specification does not require rewriting of upstream responses to filter out parts for which the queried server is not authoritative. This means that a downstream caching resolver will tend to poison its cache if it adds data from such responses that are not directly in response to the QNAME." http://seclists.org/oss-sec/2014/q4/602 http://seclists.org/oss-sec/2014/q4/602 Adding entries (including additional records, which are common in DNS) without validation is just asking for a DNS poisoning attack.
ajross 12y ago
Won't happen. It's turning ugly (really the ship sailed on that with the container stuff), but it's never going to "go off the rails". It's quality software and it works. I... guess I'm sort of in the "anti-systemd" crowd, but in the mildest sense. For the overwhelming majority of use cases, it really just isn't appreciably better. It's better in lots of measurable, small, and ultimately more or less irrelevant ways. Runtime system software maintenance just wasn't a problem worth solving, IMHO. But... it's quality software with solid features and it works. The same logic says we should just shut up and take it. So I do.
GFK_of_xmaspast 12y ago
I don't have any dogs in the systemd fight, but it looks like you're moving from 'anger' to 'bargaining' in Kuebler-Ross.

by erhardm 12y ago

UnoriginalGuy 12y ago

This was added in 2012.

erhardm 12y ago
You mean the article or the feature? The article is dated by 22/11/2014. The feature may have been added a few months/years ago, but the point still stands about systemd showing signs of feature creep. Note also the DNS resolver issue[0] [0] - http://seclists.org/oss-sec/2014/q4/592 http://seclists.org/oss-sec/2014/q4/592
- the_mitsuhiko 12y ago
  The feature was added in 2012 or something, whenever the log support came. The DNS issue affects not just systemd but any stub resolver.
  - pdkl95 12y ago
    The DNS issue is worth in systemd's resolver because it is NOT merely a "stub resolver" - it implements a cache. From the same thread: "The DNS specification does not require rewriting of upstream responses to filter out parts for which the queried server is not authoritative. This means that a downstream caching resolver will tend to poison its cache if it adds data from such responses that are not directly in response to the QNAME." http://seclists.org/oss-sec/2014/q4/602 http://seclists.org/oss-sec/2014/q4/602 Adding entries (including additional records, which are common in DNS) without validation is just asking for a DNS poisoning attack.

ajross 12y ago

Won't happen. It's turning ugly (really the ship sailed on that with the container stuff), but it's never going to "go off the rails". It's quality software and it works. I... guess I'm sort of in the "anti-systemd" crowd, but in the mildest sense. For the overwhelming majority of use cases, it really just isn't appreciably better. It's better in lots of measurable, small, and ultimately more or less irrelevant ways. Runtime system software maintenance just wasn't a problem worth solving, IMHO. But... it's quality software with solid features and it works. The same logic says we should just shut up and take it. So I do.

GFK_of_xmaspast 12y ago

I don't have any dogs in the systemd fight, but it looks like you're moving from 'anger' to 'bargaining' in Kuebler-Ross.