I’d rather be subscribed to insane immature list and have Linux itself kept sane and mature (as in my humble opinion systemd, firewalld, is continuing insanity of linux transformation. But yet, again, these are just insane thoughts of immature person: myself ;-)
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
*Including* people taking up a collection to hire a hitman. Joke or not, that’s some pretty messed up and sick stuff.
Robert Arkiletian wrote:
I just skimmed it. Gee, a lot of the open source community are assholes!
Look at all these nasty things.
For someone who read it carefully, did he say *anything* about “I don’t really understand why these attacks”, much less, “I wonder if there are a large percentage of the open source community who dislike my approach, disagree with it strongly, and feel that my way is being forced down their throats without their have the F/OSS *choice* of saying “no!””
mark
OK, I for one am boycotting his creature, to the extent I can, the way I
can: I move my servers away from Linux (to FreeBSD, if someone interested). None of my server will run on the box that has systemd. Workstations stay Linux…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
That is fair and fully your choice.
Wishing the man harm though, as some have done, is entirely irrational.
But oddly, he didn’t even mention that there would be a real simple solution – just add backwards-compatible improvements instead of actively wrecking the interfaces everyone else had depended on for decades.
When someone is saying “they are looking into hiring hitman for me” I’m not only questioning sanity of “them” (and we know many examples when majority is nasty), but also sanity of person who claims that. I do have
(my humble, yet awfully strong) opinion about mental abilities of authors of systemd. And I personally will not wish anybody of them harm (even though their systemd did a lot of harm to Linux), I just find better system for what I do…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
Digimer wrote:
Very, very old joke: a greenhorn wants to go prospecting for gold. The old muleskinner rents him a pack mule. Three times he ties a pack to the mule, loads his stuff on, and three times the mule exhales and everything falls off. Finally, he takes the mule, and drags his stuff, back to the old muleskinner, and complains, “I thought you told me this mule was easy-going, and would do what I told him!
The old muleskinner looks around, picks up a stout branch, whirls around, and *WHACKS* the mule right between the eyes. He then picks up the greenhorn’s stuff and loads it on the mule with no problem.
“But, but…” the greenhorn stutters, “I thought you told me he’d do what I wanted!”
The old muleskinner replied, “He will… but sometimes, you just have to get his attention.”
I want to *WHACK* Poettering between the eyes, and get his attention. Maybe that would get past his self-centered self-importance.
mark
you’ve said this about 8 times now.
I’m pretty sure FreeBSD has its own mailling lists.
look in the mirror.
“decades”. That, by itself, already calls for an update, no?
But so did other systems, but they later found out that sometimes you have to break this backwards to infinity compatibility in order to get some big progress.
There is even a name for this break up, and they call it “disruptive events”, “disruptive technology”, etc. When we have such events, you either get up to speed, change your market field or.. get rusty…
Sorry man, that’s how it works, everywhere. Although many will probably just “miss the old days”.. yeah..
Like for firewalld and systemd, as they were already mentioned in here. It’s hard _just because_ it’s different. But wait, wasn’t iptables different from ipchains? And is nftables going to be as the same as iptables? No, of course not. There are features in nftables that you can’t put into iptables cleanly, so you need a new workflow on it.
Marcelo
John R Pierce wrote:
Why? In what way am I coming up with new complicated overlays for what we both use, and effectively forcing you to use it?
And note that I used the word “forcing” because most organizations of any size are about as likely to *completely* change o/s’s to another vendor as they are to celebrate May Day by running through the streets in the nude.
mark
I was commenting re your constant complaining that you don’t like change, therefore its all wrong/bad/whatever. In YOUR opinion.
we get it already, give it a rest. If you want to complain to someone for choosing to use systemd or whatever, bitch at Red Hat, not CentOS, and certainly not the CentOS user community (this list).
Marcelo Ricardo Leitner wrote:
Why? Do you ride a bicycle differently, or drive differently, than you did say, 20 years ago? You went out and bought a recumbent, or an electric car?
While we’re at it, can you tell me how much better a brand new microwave, with 20 touch-buttons for misguessing how long to cook something, is better than the old microwave I used to have that had a “cook/defrost”
dial, and a timer dial?
Actually, I struggled with ipchains, and found iptables much simpler. I’ve yet to see anyone suggest that systemd is “simpler”.
There was much wailing a gnashing of teeth from “purists” when these things came in. “But it’s not really driving!” some would say. “It lets people be lazy!” others would say. Many of those people still drive old cars, such is their choice.
Today, overall, the roads are much safer. Change is good.
John R Pierce wrote:
vendor
First, I refute that – I do not “constantly complain” that I don’t like change; I do have every right to join others, with all of us complaining that we dislike a significant sweeping change. And, for that matter, most of us don’t have time in our life to join the fedora developer’s list, and argue there long enough and loudly enough to promote change.
Does this also apply to everyone else on the list complaining? Why single me out, esp. since I was *not* the OP for this thread.
mark
No, do you dig a new foundation for your house every 10 years? Trade in your wife and kids?
Only if the design was bad in the first place. And if the design was really bad, there wouldn’t be any users to infuriate by breaking the interfaces they use. But the unix design that linux and linux distributions copied was pretty good, including the way init started things.
I doesn’t have to be that way. But with free software when it breaks you get to keep all the pieces.
Not sure iptables ever got it right in the first place. No one to copy from…
I think I’ve forgotten what user interfaces these break. Did they take away the steering wheel to add them?
The complaining about systemd (again) started with your comment:
===Can someone just send the team that’s working systemd on a nice vacation, say, maybe northern Iraq/Syria, the land of ISIS?
===
You advocated indirectly for a contributor’s harm because you don’t like his work. Before that, Steve Clark referenced the /. article and Jonathan commented on his preference for Phronix as a news source.
You cropped out the second half of my analogy, which explained why I
thought the analogy applied.
Umm, no. They are pretty much backwards compatible. Otherwise you’d have to go through new training and testing to be certified as a driver when you buy a new car. And car vendors know better than to do that to their customers.
Well, yes. Bit off topic now, but yes. The core way of doing it, is the same, but nowadays if I don’t wear proper clothing and pay extra attention, I’ll probably be overrun in the next corner. It changed.. I
can’t ride it just like before.
Heh.. so let’s all use mud jars instead of plastic ones, or ceramic pans instead of the nonadherent ones we have nowadays. ;)
Ok, that aside, sorry, I’m not saying that having 20 touch-buttons is right/better, though. If you don’t like it, don’t buy it.. But the old dial is gone because (some) people needed something else, just that… My dial used to break now and then, btw, and touch buttons are easier to clean up, IMO. Mine has a grill feature that I never used for ~7 years. When I tried to use, it burnt and left the food smelling melted plastic.
;D IOW, no good for me, at least for now. But if I find a recipe that needs it, perhaps I’ll fix my microwave just to try it.
Heh, me too, but well, 20 buttons is more complex than 1 dial, yet more practical if you know how to use them.
Marcelo
they took away the clutch pedal.
anyways, cars are not a good analogy to computers over the same time scale, unless you want to go back to the days of the model T, where the
3 pedals operated clutch bands on a planetary transmission, and the throttle and ignition timing were levers on the steering wheel, and the brakes were a hand lever.
computers have evolved far faster than automobiles over the last 40
years that I’ve been in this industry. maybe I should start whining about lower case, and these damn interactive guis, after all hollerith punchcards and batch processing was good enough in the 1970s! Why, we could get amazing stuff done with 8K words of core, and a 1000K word hard disk.
why should mysqld need to wait for ntpd to start, and ntpd wait for sshd to start, and sshd wait for cups? multiply this by the 30 or so services running on a typical server. sysVinit is a crude hack that has no concept of service dependencies.
anyone remember when inittab was *it*, and sysVinit came out, and EVERYONE COMPLAINED ABOUT HOW COMPLICATED IT WAS?
god, you whiners are a joke. go back to slackware 0.9.6 circa 1994 (20
years ago, I believe that was your threshold) and see if you even remember how to configure anything.
I’d compare those to the pre-sysV unix system designs. Where sysV
became the standard to follow – or copy pretty explicitly like linux distributions did. Which was why we used them.
So now we have hardware hundreds of times faster, and you are trying to tell me it can’t do the same thing in a backward compatible way???
Really? Are you really comparing this to technology stuff?
Anyway, hands down if you still use one of the very first mobile phones and not a smartphone, or if your laptop is 10, 20 years old.
That’s very subjective. There were those who loved LPRng and those who couldn’t bare with it, which saw salvation on CUPS (like me).
The only thing in common between both was the lpr/q commands and that they “manage printing”, because everything else was different. Today LPRng is long gone..
It’s not saying that it is “bad”, but that we can do more, and for that we can’t have all that backward compatibility altogether. It’s already very hard to QA just the new bits, imagine with everything combined. IMHO better do one way and do it good.
Marcelo
But the basic operation staid the same – brake on left, gas on the right, gear shift lever, steering wheel, etc, etc.
Yes, wasn’t it amazing how much could get done with so little resources. We ran our whole college administration on an IBM-1130 with 8K of core and a 2.5mega byte removable drive.
And now the font rendering takes more computing power than all the math you did to come up with what you wanted to print. Progress.
Look at it from this perspective: assume your bank and retirement accounts and any stock transactions you have are managed by very complex software sitting on top of the interfaces these distributions provide. Now how wild and crazy do you want the changes in those interfaces to be? How much do you want the fees you pay to increase just to keep the same applications running with the same amounts in the accounts as they showed before?
I use a modern clamshell/flipphone. fits in my pocket much better than a smartphone and is better for making phone calls.
John R Pierce wrote:
$(x=1)++
Same here – my main use of the device is as a cellular “telephone”, so that I can speak to someone at a distance , and I really *don’t* want to “entertain” say, half the subway car with my conversation, as I speak that loudly so as to be heard when the mike’s not at my mouth….
And I’m really coming to *loathe* smartphones, and the people who can’t put them down to drive, or see the light change, or deal with the cashier in the supermarket, or….
mark
Sorry about that. I probably will be in minority – the ones who would prefer childishness and insanity on the list in hope to push opinions towards more sanity in Linux distribution(s) we do (did?) love for ages…
Indeed, and I’m on [some of] them. And I will not hesitate to post _there_
that “this” IMHO is done better by Linux folks when I find something that is better done in one or another Linux distro.
My apologies again.
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
No. I got my portion too. For consistently mentioning FreBSD. So, you are not alone. But I’ve apologized. And, hey, didn’t I per chance get my portion from you? Never mind, whomever from, I deserved it ;-)
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
Moving the VT management into userspace. Having it in-kernel makes it
1.) difficult to program for (and hasn’t been updated much since the 90s)
2.) insecure and
3.) better support multi-seat environments
amongst other reasons. BTW, this is all code by someone other than LP, so add Mr. Herrmann to your conspiracy files.
Errr… I meant that moving it to userspace makes it easier to support multi-seat environments.
In your opinion; clearly John (and I) disagree.
I can’t speak for John, but presumably you were singled out for making your complaint in a completely ridiculous and inappropriate way.
–keith
Who are you and who is
Mods?
Please take this pissing contest off-list if you would all be so kind.
John This is all happening because my father didn’t buy me a train set as a kid.
— Warren Buffett, joking about his decision to buy a railroad, the Burlington
Northern Santa Fe Corporation, New York Times, 4 November 2009
Just wait
Opps i dont have ssc yet who was that over
Cheers,
Cliff
Am 08.10.2014 um 19:35 schrieb Valeri Galtsev :
I wonder why personal implications are communicated without explained reasons
(because all are screaming doesn’t mean its valid). “Changes” have inherent implications by it self – it doesn’t matter what was the change. Take a look back (or step) and try to see what is being ignored, changes all over the community projects (e.g. kernel, gui, distros, wm, nfs, config syntax). So whats the point and why bothers other, if the problem [1] lies in our selfs? BTW, you have the choice.
[1] problem definition: actual state = EL6, goal = EL7 and something that
is preventing the transition (and that is for sure nothing technical).
PS: I’m not taking a position for systemd – just seeing some kind of structure.
Hi Jonathan,
What exactly does that mean – multi seat environments?
I have changed from free gear to fixed gear – hell, that was terrible on the first days. I am still riding the fixed one.
Am 08.10.2014 um 20:25 schrieb Les Mikesell :
was – the requirements at that time were nearly/completely different. We have different scenarios right now.
Really? What application could you not start with sysv init syntax?
What CPU has become too slow to start things serially? What feature do you need that could not have been added without breaking other existing work?
The feature of advantage is fast boot. As Linux like Windows needs reboot often, it is awfully important. And all of you, dinosaurs (who saw years long uptime of Linux machines) who don’t care that boot takes 60 seconds now instead of 4 minutes should just shut up.
Let me second what you said. I also would add: In my opinion it is not clever to keep settings that are expressed by plain ASCII text being marked up, “dressed into junk”, XML. For human to read them you need
“undress” them (you GUI guys may forget that your GUI does that – not literally of course), and to pass them to systemd itself one has strip the junk (XML markup). The same goes about firewalld.
But what am I doing. The World passed that point…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
I guess debugging the GUIs that make the config files accessible will be job security for the young guys that replace us…
Are you serious? Do you provide manually the links for the sysV fs?
Or do you use an “wrapper/helper” (e.g. ntsysv, chkconfig)? So why not using the same work-abstraction for a different boot process?
And about readability – do you read all configuration files on every system boot? No – but the system does and therefore xml is more suitable.
Please don’t misunderstand me. As I tried and mentioned before – the whole disarmony here is more based on human and not on technical factors. I am on your side but for sure for different reasons. EL7 has changed to much not only the boot process. The cluster stack, IPsec implementation and httpd config syntax is different, NM for network, gnome-shell, missing TB and so on. Further investigation and adaptation of our deployment processes are in progress. So, EL6 will stay here as a primary plattform.
Peace.
I have. And I know how to. Ascii sort order is a straightforward concept for both humans and computers. I knew how to deal with scenarios with subtle issues like the network service ‘completing’
startup even though the connected switch was still doing spanning tree and the next network operation would fail. And now I don’t know how to tell how introducing some new arbitrary service will interact with an existing arbitrary set.
I have. And now those won’t work the same way. So why was using that abstraction good?
I read them at the important times. When things aren’t working. The ones shipped with the system have already had years of breakage gradually fixed in fedora. Our local apps will have to start from scratch.
You are just deferring the pain unless you plan to retire before EOL
for EL6 and foist the work off on someone else.
Not that simple. Smaller things none of which on its own make you make fundamental decision just accumulate, so after some critical mass you are there, the decision is made. I probably will not remember all of them, and definitely not in the order of their arrival.
1. SELinux (it doesn’t not pass _my_ “security has been enhanced”
estimate. It can be turned off on the fly, therefore it is as if it doesn’t exist. It adds tens of thousands of lines of code to the kernel, thus increasing chance of bugs and deteriorating security. And once my estimate was confirmed: “critical security update of SELinux”…)
2. The need to reboot linux box often (on average every 45 days in my observation…). I know, there are workarounds to patch on the fly…
3. Firewalld. I’m tempted to say: the philosophy it’s based on is flawed. Hey, this is our institutional internal network which is behind strict firewall, so we consider it “safe” zone… Nope. Wherever is the machine that _you_ do not administer, you better do not consider it “safe” or
“secure”. It uses the same iptables kernel module, right? And you can switch to using iptables instead if it, right? Only one day you end up running distro with stripped this, stripped that… Small thing in itself as I said.
4. Systemd. (“If it ain’t broke, don’t fix it”). I don’t want to start flames again…
And things add up gradually till you realize this. If I were willing to run M$ Windows I would pay for it, and will curse it legitimately as I
paid for it and my money entitles me to curse it. Now I run free open source system courtesy of uncounted number of generous developers. I
should be grateful to them and tolerate new routes they take – which is usually their only reward for their work… So, when I can not tolerate it any more (… accumulated above critical mass), I just go to another also open source system, thank goodness, they still do exist…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
In this phrase I was always wondering “as a kid” part is about “me” or about “my father” ;-)
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
Car vendors might behave differently however. Particularly if they obtained revenue from the necessary retraining and profited from the licensing procedures.
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
Ok I read the information. So as I understand it you are going have a computer that has multiple graphics cards with multiple keyboards and multiple mice divided into seats. Really?
Where do I buy this computer?
Yes, “much too cavalier about bugs and compatibility” is exactly right. Never mind the fact that you break everything you existing users have done, you might attract some people who like windows and don’t understand that it took decades of patches and reboots every Tuesday to reach any kind of stability with its monolithic approach.
Once upon a time, Steve Clark said:
Umm, Wal-Mart? Best Buy? You do realize that AMD/ATI Radeon chips (for example) in many systems can drive multiple display simultaneously? Add another USB keyboard/mouse/speakers, and you can set them up as multiple seats. Pretty much any modern PC can have more than one video card.
It is much simpler to run remote X sessions over a network for multiuser access and probably not much more expensive if you use older PCs as terminals. You do have to boot something, but x2go or freenx/NX are cross platform and have great remote performance. I’m surprised no one has made a mini-linux distro that boots straight to x2go for this purpose, but if they have, I haven’t found it.
Or a raspberry-pi for $50 if you don’t have an old PC.
Am 09.10.2014 um 16:57 schrieb Les Mikesell :
As I said, because you don’t known doesn’t mean its faulty -> human factor.
An interesting idea, that was born in your head, right?
I would suggest to keeping the facts on the table and concatenating opinions and other nonsense to device null.
“human factor” means it is extra work for someone who was already using a linux distribution because it did what they wanted – which doesn’t seem good to me either. And since I don’t understand it, I
don’t know how to prove that some arbitrary addition can’t cause a dependency loop when added to some arbitrary existing configuration. Do you?
No, it is a very real scenario for me. If I leave the upgrades for someone else it will probably be the end of linux use here.
The harsh reality of life is ‘real’ villains etc. don’t talk about harming someone else, they just do it or employ others to do it. Talking, discussions, proposals etc. are dangerous to the instigators because it creates implicating criminal evidence.
Systemd should be been much more widely discussed with the general Linux / Red Hat and clones community. I am sure beneficial improvements would have been proposed and, hopefully, implemented prior to Red Hat’s adoption.
That’s life. Few things are perfect :-)
Change is inevitable in life. Virtually everything changes including the eventual decline of our sun. Not sure about atomic weights or the value of Pi (3.142) or E=mc² Perhaps they break the rule that everything changes.
Being optimistic perhaps systemd will quickly change into something more welcoming to the vast majority of users.
—
Regards,
Paul. England, EU.
It is about fundamental approach. We always modularize things: split into smaller subunits each of the last doing its smaller task. This allows to make smaller things work reliably, and test these smaller things more comprehensively. As it is much smaller number of combinations of factors you need to repeat your test with in case of subunits. People use this approach for ages. Programs are split into subroutines. Rockets are built from to awful degree independent modules. We had this “modular” system V
boot until recently. We lost it. We got “iPhone, with whatever you can get in App store” instead. And not all of us are pleased by this change. And, BTW, there was one of the posts of MS Windows big fan on this list who welcomes this change; his post should have made everybody think…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
As a computer person of nearly 50 years experience (wish I was 20
again), I have long recognised the essential fact ….
* That if one gets it Right the first time, one inevitably saves an awful lot of time and an awful lot of effort subsequently trying to get it Right or even to “improve” the less than perfect item.
A little critical and constructive effort into the architecture and functionality of systemd would have brought considerable benefit. Perhaps the Fedora and RedHat persons might remember this advice when the next “wonder” is proposed. No reason not to want *genuine*
perfection for all.
Astonishing ! Its how I do my projects, programming and other work. Do you think our Fedora friends might adopt your revolutionary idea ?
Well, one suggestion is we set-up a working group to devise a replacement (or replacements) for systemd. The most difficult task is selecting a nice name for the systemd replacement.
No good anyone moaning when only action will bring change.
No. I think they see the huge market that Windows has and want to emulate their approach at the expense of anyone who liked unix and headless multi-user servers. But the people who like windows don’t really need a bad imitation either.
I’m curious on how you think that SysV init is modular in this approach and systemd isn’t?
A simple ‘ps uf -p 1’ on a couple of machines shows about 10x the resident memory use and 5x virtual on CentOS 7 vs. 5.x. And yet, the programs that it started don’t show any improvement for the extra cost.
So, the extra memory is being used by systemd processes?
What makes you think that systemd is not modular? Have you actually looked at its structure (let alone the code)? If you look inside /usr/lib/systemd/ do you see one big monolithic library which represents one big failure point, or do you see a few dozen dedicated small libraries, each doing one particular thing?
I don’t really see how systemd violates the “do one thing and do it well” philosophy. A lot of people seem to ignore the fact that systemd is *not* one binary executable which replaces init and tries to take control of everything, but rather a whole swarm of independent binaries, each in charge of one particular function of the OS. If one of them breaks for some reason, others will still keep functioning.
It appears to me that much of bashing of systemd is just FUD. One of the typical misconceptions is the disable vs. mask for services –
+1, , and whatever else is in vogue.
hear here! (or is that, here hear?)
Once upon a time, Marko Vojinovic said:
systemd (as PID 1) is not necessarily the problem. The problem IMHO is the systemd _project_ that appears to have a severe case of scope creep. They have swallowed up other projects (udev, dbus), reinvented wheels
(logging, ntp, network configuration), and appear to still be growing without bounds. When people don’t like some of the decisions of systemd developers, and the systemd project keeps taking over more of the core OS functionality, it is frustrating to watch.
I don’t quite disagree, but: a number of programs in Fedora/RHEL/CentOS
actually use (or possibly abuse) chkconfig to see if a service should be run from cron or by some other trigger. The systemd implementation is more clean about being used for “disable starting at boot”.
I agree that it ended up being different terminology than I think is most clear and ideal, but… c’mon, we’re sysadmins. If we start listing all of the things in Linux with that problem we will be here all day. Eventually, you just learn it, think, “whelp, another one for the that’s-a-little-weird file”, and move on.
There’s probably nothing wrong with these things in the context of a new/different OS oriented to servicing a single user who is assumed to own the world as a side effect of logging into a magical console device and starting a GUI where things can pop into existence and chat with other things easily. But it doesn’t have much to do with unix-like concepts.
What happens to ownership of a DVD or audio device when a different user logs in at the console – even if some other remote user wants to access them? The magic is more about ConsoleKit and PolicyKit than specifically systemd but the underlying dbus gunk seems to be merging together.
And the monthly reminder:
Please watch your replies to this list. Reindl is _not_ a subscriber and all you do is inject his diatribes back here where he’s not wanted thus circumventing his removal.
Thank you.
John
You think that nobody on that project thought about this before?
It’s not just remote X sessions. You want at least USB and audio redirection and also a decent 3D performance.
We currently do that using spice for VMs, I don’t know how feasible it is to run it on a real hardware.
There are some good pro’s on this setup:
– this installation is physically simpler than having 4 full computers as it requires 1/4 of the wall plugs and network points
– no single point of failure (as in: 4 seats down is okay), if you compare with ones using x2go and similar (application server)
– easily scalable: need more seats? buy 1 computer more, you have +4
seats, and you’re good. No server needs to be re-evaluated.
– easier to maintain, as you maintain 1/4 of the systems you would otherwise.
– very cost effective with commodity hardware, that everyone knows how to deal with.
– vendor independent
And probably many others that I forgot :)
Not saying it’s the best, though. Just saying that yes this is a good project that is well plotted and has its audience.
Marcelo
Yes but you have to be physically close to the main cpu. What about distractions from other people sitting right next to you?
Playing music, etc.
On 10/13/2014 7:17 AM, Steve Clark wrote:> Yes but you have to be physically close to the main cpu. What about
That’s not all that different from modern cube farms. You learn to tolerate or ignore other people, or more ideally, collaborate well with your closest co-workers.
Where I work, there are people sitting side-by-side at folding tables (business has picked up faster than physical facilities can keep up with). In our case, they’re all using zero clients and virtual desktops. However, it’s exactly the kind of setup where a multi-seat computer might make sense for other companies or schools that don’t have our virtualization expertise.
Being able to grab your existing desktop remotely with all open windows and long-running programs intact is a big plus, though – and you get that for free with NX or x2go. Can you connect remotely to your VM host some other way if you aren’t at the special desktop?
How much different is that from VNC? Just curious.
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
VNC just does a basic bit-copy of the screen for the remote side so it is not particularly efficient, although it might be better if you are mirroring a real console instead of running a virtual session set up by VNCserver. NX/x2go have a full proxy/stub X server and client at both ends with tunable caching/compressiion on the remote side.so things like font rendering and block moves for scrolling are much faster. X2go also can map audio/drives/printers if you want.
Both are packaged and fairly easy to try on CentOS 6. On 7, only x2go is available and it has a problem with the 3d requirement of Gnome3 so you have to use KDE or install MATE from EPEL.
I didn’t mean to imply that you use a multi-seat computer to get to a desktop served by a remote machine, though you could certainly do that if you wanted. Everyone still needs a machine to function as an endpoint for the remote desktop to be delivered to, though. You use a multi-seat computer when you don’t have enough computers to give everyone their own machine.
Like William Gibson said, “The future is already here – it’s just not evenly distributed.” I have enough computers that I could make furniture out of them, but I’m sure there’s some cash-strapped school district using donated hardware that would jump at the chance to have their computer lab serve ten students at a time instead of five.
— -Chris
I haven’t used x2go, but NX is way faster than VNC over a slow link
(e.g., home DSL, hotel wifi) in my experience.
–keith
Ah, just remembered one thing. If memory serves, FreeNX dates back to the same time that multiseat project was started. They were developed targeting a very similar use case, but different in the end, while multiseat had a faster time-to-market, I think.
Places like some public libraries, like we have here in Brazil, uses multiseat. Keeping these running is much easier than having someone with the knowledge to fix a virtualization host when it goes rogue. These places here have 0 tech people working in there..
Note that multiseat doesn’t even require a network at all. You can always open up a text editor and do some writing, or some spreadsheet calcs…
It’s been a while since I last used FreeNX. I didn’t recall it being able to map drivers back then but good to know that x2go can do it, thanks. Back then it already was indeed much faster than VNC and/or remote X.
Marcelo
The K12LTSP project might have been a good fit several years ago. This was a respin of fedora or CentOS 4/5 distributions that would come ready to PXE-boot diskless PCs as thin clients and host their sessions. I think you could make them auto-login to a kisok type application if you didn’t want individual user logins. That involved more hardware, of course, but a lot of places used old donated boxes with the disks removed and only needed one real server per classroom of 30 or so. The project evolved (or devolved, depending on how you look at it) into packages that you have to install on a stock system and that now have more of a fat-client with local apps approach
(k12osn).
Sure, but networks are cheap and reliable and easy to run longer distances than vga cables.
On a one-hop network, even straight remote X is reasonable for thin-type clients. The main problem remaining is doing full resolution streaming video to many desktops from a multiuser host –
which the multi-video card approach should do for some number of seats.
But, I’m kind of surprised that someone hasn’t done a raspberry-pi type device that boots directly into x2go and comes out cheaper than a video card per seat. Haven’t needed one badly enough to build it myself yet.
It should be trivial to set up an actual RPi to do that.
One of the options during installation is to boot into text mode instead of graphical mode. Once it
Not really. The beauty of the original K12LTSP respin was that just you did a normal fill-in-the-form install pretty much like any fedora/CentOS in a server with 2 NICs and you could plug in some diskless PCs and they came up working with applications ready to go. Compared to that, there’s still a lot of do-it-yourself assembly required. I recall from the early K12LTSP mail list days that quite a few teachers were able to set up a classroom with very little admin knowledge.
Agreed on the hardware front, but couldn’t this be a canned image you copy to an SD card with some way to edit the target IP address in place without needing to rebuild it?
If you insist on having a whole OS dedicated to this, I guess you could go fork Raspbian (http://www.raspbian.org/) and add this stuff to the installer.
Compare that amount of work to:
1. Install Raspbian from NOOBS
2. Change boot option so it doesn
I’m not insisting on anything, I just don’t see a typical classroom teacher doing that – in addition to knowing what to set up on the server side to match. Where they could just follow some simple instructions with the LTSP install and come up working.
Agreed, but K12LTSP worked mostly as-is for a lot of districts.
I don’t need it myself – I’m just surprised that someone hasn’t done it already in a way that would be useful without everyone having to start from scratch themselves. And maybe that’s not even the best approach – maybe a ‘fat’ client with NFS-mounted or cloud-like storage would be better if there is an automated way to keep it up to date without letting the user modify things. I guess these days you’d have to balance the cost of administering the things against buying a chromebook.
99 thoughts on - Systemd Adding Its Own Console To Linux Systems
I’d avoid slashdot entirely and read the phoronix article:
http://www.phoronix.com/scan.php?page=news_item&px=MTgwNzQ
Also, note that this is in the experimental release and most likely won’t be EL7’s systemd any time soon.
Jonathan Billings wrote:
And the point of it is?
Can someone just send the team that’s working systemd on a nice vacation, say, maybe northern Iraq/Syria, the land of ISIS?
mark
Really? Really?
Can we please keep this list sane and mature? That’d be great.
Just a heads up to those who haven’t seen this yet. The main author of systemd publicly wrote about being basically persecuted.
https://plus.google.com/u/0/+LennartPoetteringTheOneAndOnly/posts/J2TZrTvu7vd
I’d rather be subscribed to insane immature list and have Linux itself kept sane and mature (as in my humble opinion systemd, firewalld, is continuing insanity of linux transformation. But yet, again, these are just insane thoughts of immature person: myself ;-)
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
*Including* people taking up a collection to hire a hitman. Joke or not, that’s some pretty messed up and sick stuff.
Robert Arkiletian wrote:
I just skimmed it. Gee, a lot of the open source community are assholes!
Look at all these nasty things.
For someone who read it carefully, did he say *anything* about “I don’t really understand why these attacks”, much less, “I wonder if there are a large percentage of the open source community who dislike my approach, disagree with it strongly, and feel that my way is being forced down their throats without their have the F/OSS *choice* of saying “no!””
mark
OK, I for one am boycotting his creature, to the extent I can, the way I
can: I move my servers away from Linux (to FreeBSD, if someone interested). None of my server will run on the box that has systemd. Workstations stay Linux…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
That is fair and fully your choice.
Wishing the man harm though, as some have done, is entirely irrational.
But oddly, he didn’t even mention that there would be a real simple solution – just add backwards-compatible improvements instead of actively wrecking the interfaces everyone else had depended on for decades.
When someone is saying “they are looking into hiring hitman for me” I’m not only questioning sanity of “them” (and we know many examples when majority is nasty), but also sanity of person who claims that. I do have
(my humble, yet awfully strong) opinion about mental abilities of authors of systemd. And I personally will not wish anybody of them harm (even though their systemd did a lot of harm to Linux), I just find better system for what I do…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
Digimer wrote:
Very, very old joke: a greenhorn wants to go prospecting for gold. The old muleskinner rents him a pack mule. Three times he ties a pack to the mule, loads his stuff on, and three times the mule exhales and everything falls off. Finally, he takes the mule, and drags his stuff, back to the old muleskinner, and complains, “I thought you told me this mule was easy-going, and would do what I told him!
The old muleskinner looks around, picks up a stout branch, whirls around, and *WHACKS* the mule right between the eyes. He then picks up the greenhorn’s stuff and loads it on the mule with no problem.
“But, but…” the greenhorn stutters, “I thought you told me he’d do what I wanted!”
The old muleskinner replied, “He will… but sometimes, you just have to get his attention.”
I want to *WHACK* Poettering between the eyes, and get his attention. Maybe that would get past his self-centered self-importance.
mark
you’ve said this about 8 times now.
I’m pretty sure FreeBSD has its own mailling lists.
look in the mirror.
“decades”. That, by itself, already calls for an update, no?
But so did other systems, but they later found out that sometimes you have to break this backwards to infinity compatibility in order to get some big progress.
There is even a name for this break up, and they call it “disruptive events”, “disruptive technology”, etc. When we have such events, you either get up to speed, change your market field or.. get rusty…
Sorry man, that’s how it works, everywhere. Although many will probably just “miss the old days”.. yeah..
Like for firewalld and systemd, as they were already mentioned in here. It’s hard _just because_ it’s different. But wait, wasn’t iptables different from ipchains? And is nftables going to be as the same as iptables? No, of course not. There are features in nftables that you can’t put into iptables cleanly, so you need a new workflow on it.
Marcelo
John R Pierce wrote:
Why? In what way am I coming up with new complicated overlays for what we both use, and effectively forcing you to use it?
And note that I used the word “forcing” because most organizations of any size are about as likely to *completely* change o/s’s to another vendor as they are to celebrate May Day by running through the streets in the nude.
mark
I was commenting re your constant complaining that you don’t like change, therefore its all wrong/bad/whatever. In YOUR opinion.
we get it already, give it a rest. If you want to complain to someone for choosing to use systemd or whatever, bitch at Red Hat, not CentOS, and certainly not the CentOS user community (this list).
Marcelo Ricardo Leitner wrote:
Why? Do you ride a bicycle differently, or drive differently, than you did say, 20 years ago? You went out and bought a recumbent, or an electric car?
While we’re at it, can you tell me how much better a brand new microwave, with 20 touch-buttons for misguessing how long to cook something, is better than the old microwave I used to have that had a “cook/defrost”
dial, and a timer dial?
Actually, I struggled with ipchains, and found iptables much simpler. I’ve yet to see anyone suggest that systemd is “simpler”.
mark
Airbags, ABS, Traction Control, ACE compatibility, stronger survival space, better fuel economy, more comfortable…
There was much wailing a gnashing of teeth from “purists” when these things came in. “But it’s not really driving!” some would say. “It lets people be lazy!” others would say. Many of those people still drive old cars, such is their choice.
Today, overall, the roads are much safer. Change is good.
John R Pierce wrote:
vendor
First, I refute that – I do not “constantly complain” that I don’t like change; I do have every right to join others, with all of us complaining that we dislike a significant sweeping change. And, for that matter, most of us don’t have time in our life to join the fedora developer’s list, and argue there long enough and loudly enough to promote change.
Does this also apply to everyone else on the list complaining? Why single me out, esp. since I was *not* the OP for this thread.
mark
No, do you dig a new foundation for your house every 10 years? Trade in your wife and kids?
Only if the design was bad in the first place. And if the design was really bad, there wouldn’t be any users to infuriate by breaking the interfaces they use. But the unix design that linux and linux distributions copied was pretty good, including the way init started things.
I doesn’t have to be that way. But with free software when it breaks you get to keep all the pieces.
Not sure iptables ever got it right in the first place. No one to copy from…
I think I’ve forgotten what user interfaces these break. Did they take away the steering wheel to add them?
The complaining about systemd (again) started with your comment:
===Can someone just send the team that’s working systemd on a nice vacation, say, maybe northern Iraq/Syria, the land of ISIS?
===
You advocated indirectly for a contributor’s harm because you don’t like his work. Before that, Steve Clark referenced the /. article and Jonathan commented on his preference for Phronix as a news source.
You cropped out the second half of my analogy, which explained why I
thought the analogy applied.
Umm, no. They are pretty much backwards compatible. Otherwise you’d have to go through new training and testing to be certified as a driver when you buy a new car. And car vendors know better than to do that to their customers.
Well, yes. Bit off topic now, but yes. The core way of doing it, is the same, but nowadays if I don’t wear proper clothing and pay extra attention, I’ll probably be overrun in the next corner. It changed.. I
can’t ride it just like before.
Heh.. so let’s all use mud jars instead of plastic ones, or ceramic pans instead of the nonadherent ones we have nowadays. ;)
Ok, that aside, sorry, I’m not saying that having 20 touch-buttons is right/better, though. If you don’t like it, don’t buy it.. But the old dial is gone because (some) people needed something else, just that… My dial used to break now and then, btw, and touch buttons are easier to clean up, IMO. Mine has a grill feature that I never used for ~7 years. When I tried to use, it burnt and left the food smelling melted plastic.
;D IOW, no good for me, at least for now. But if I find a recipe that needs it, perhaps I’ll fix my microwave just to try it.
Heh, me too, but well, 20 buttons is more complex than 1 dial, yet more practical if you know how to use them.
Marcelo
they took away the clutch pedal.
anyways, cars are not a good analogy to computers over the same time scale, unless you want to go back to the days of the model T, where the
3 pedals operated clutch bands on a planetary transmission, and the throttle and ignition timing were levers on the steering wheel, and the brakes were a hand lever.
computers have evolved far faster than automobiles over the last 40
years that I’ve been in this industry. maybe I should start whining about lower case, and these damn interactive guis, after all hollerith punchcards and batch processing was good enough in the 1970s! Why, we could get amazing stuff done with 8K words of core, and a 1000K word hard disk.
why should mysqld need to wait for ntpd to start, and ntpd wait for sshd to start, and sshd wait for cups? multiply this by the 30 or so services running on a typical server. sysVinit is a crude hack that has no concept of service dependencies.
anyone remember when inittab was *it*, and sysVinit came out, and EVERYONE COMPLAINED ABOUT HOW COMPLICATED IT WAS?
god, you whiners are a joke. go back to slackware 0.9.6 circa 1994 (20
years ago, I believe that was your threshold) and see if you even remember how to configure anything.
I’d compare those to the pre-sysV unix system designs. Where sysV
became the standard to follow – or copy pretty explicitly like linux distributions did. Which was why we used them.
So now we have hardware hundreds of times faster, and you are trying to tell me it can’t do the same thing in a backward compatible way???
Really? Are you really comparing this to technology stuff?
Anyway, hands down if you still use one of the very first mobile phones and not a smartphone, or if your laptop is 10, 20 years old.
That’s very subjective. There were those who loved LPRng and those who couldn’t bare with it, which saw salvation on CUPS (like me).
The only thing in common between both was the lpr/q commands and that they “manage printing”, because everything else was different. Today LPRng is long gone..
It’s not saying that it is “bad”, but that we can do more, and for that we can’t have all that backward compatibility altogether. It’s already very hard to QA just the new bits, imagine with everything combined. IMHO better do one way and do it good.
Marcelo
But the basic operation staid the same – brake on left, gas on the right, gear shift lever, steering wheel, etc, etc.
Yes, wasn’t it amazing how much could get done with so little resources. We ran our whole college administration on an IBM-1130 with 8K of core and a 2.5mega byte removable drive.
And now the font rendering takes more computing power than all the math you did to come up with what you wanted to print. Progress.
Look at it from this perspective: assume your bank and retirement accounts and any stock transactions you have are managed by very complex software sitting on top of the interfaces these distributions provide. Now how wild and crazy do you want the changes in those interfaces to be? How much do you want the fees you pay to increase just to keep the same applications running with the same amounts in the accounts as they showed before?
I use a modern clamshell/flipphone. fits in my pocket much better than a smartphone and is better for making phone calls.
John R Pierce wrote:
$(x=1)++
Same here – my main use of the device is as a cellular “telephone”, so that I can speak to someone at a distance, and I really *don’t* want to “entertain” say, half the subway car with my conversation, as I speak that loudly so as to be heard when the mike’s not at my mouth….
And I’m really coming to *loathe* smartphones, and the people who can’t put them down to drive, or see the light change, or deal with the cashier in the supermarket, or….
mark
Sorry about that. I probably will be in minority – the ones who would prefer childishness and insanity on the list in hope to push opinions towards more sanity in Linux distribution(s) we do (did?) love for ages…
Indeed, and I’m on [some of] them. And I will not hesitate to post _there_
that “this” IMHO is done better by Linux folks when I find something that is better done in one or another Linux distro.
My apologies again.
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
No. I got my portion too. For consistently mentioning FreBSD. So, you are not alone. But I’ve apologized. And, hey, didn’t I per chance get my portion from you? Never mind, whomever from, I deserved it ;-)
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
Moving the VT management into userspace. Having it in-kernel makes it
1.) difficult to program for (and hasn’t been updated much since the 90s)
2.) insecure and
3.) better support multi-seat environments
amongst other reasons. BTW, this is all code by someone other than LP, so add Mr. Herrmann to your conspiracy files.
Errr… I meant that moving it to userspace makes it easier to support multi-seat environments.
In your opinion; clearly John (and I) disagree.
I can’t speak for John, but presumably you were singled out for making your complaint in a completely ridiculous and inappropriate way.
–keith
Who are you and who is
Mods?
Please take this pissing contest off-list if you would all be so kind.
John This is all happening because my father didn’t buy me a train set as a kid.
— Warren Buffett, joking about his decision to buy a railroad, the Burlington
Northern Santa Fe Corporation, New York Times, 4 November 2009
Just wait
Opps i dont have ssc yet who was that over
Cheers,
Cliff
Am 08.10.2014 um 19:35 schrieb Valeri Galtsev:
I wonder why personal implications are communicated without explained reasons
(because all are screaming doesn’t mean its valid). “Changes” have inherent implications by it self – it doesn’t matter what was the change. Take a look back (or step) and try to see what is being ignored, changes all over the community projects (e.g. kernel, gui, distros, wm, nfs, config syntax). So whats the point and why bothers other, if the problem [1] lies in our selfs? BTW, you have the choice.
[1] problem definition: actual state = EL6, goal = EL7 and something that
is preventing the transition (and that is for sure nothing technical).
PS: I’m not taking a position for systemd – just seeing some kind of structure.
Hi Jonathan,
What exactly does that mean – multi seat environments?
Am 08.10.2014 um 20:18 schrieb m.roth@5-cent.us:
I have changed from free gear to fixed gear – hell, that was terrible on the first days. I am still riding the fixed one.
Am 08.10.2014 um 20:25 schrieb Les Mikesell:
was – the requirements at that time were nearly/completely different. We have different scenarios right now.
Really? What application could you not start with sysv init syntax?
What CPU has become too slow to start things serially? What feature do you need that could not have been added without breaking other existing work?
The feature of advantage is fast boot. As Linux like Windows needs reboot often, it is awfully important. And all of you, dinosaurs (who saw years long uptime of Linux machines) who don’t care that boot takes 60 seconds now instead of 4 minutes should just shut up.
Let me second what you said. I also would add: In my opinion it is not clever to keep settings that are expressed by plain ASCII text being marked up, “dressed into junk”, XML. For human to read them you need
“undress” them (you GUI guys may forget that your GUI does that – not literally of course), and to pass them to systemd itself one has strip the junk (XML markup). The same goes about firewalld.
But what am I doing. The World passed that point…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
I guess debugging the GUIs that make the config files accessible will be job security for the young guys that replace us…
http://www.freedesktop.org/wiki/Software/systemd/multiseat/
Am 09.10.2014 um 15:07 schrieb Les Mikesell:
Are you serious? Do you provide manually the links for the sysV fs?
Or do you use an “wrapper/helper” (e.g. ntsysv, chkconfig)? So why not using the same work-abstraction for a different boot process?
And about readability – do you read all configuration files on every system boot? No – but the system does and therefore xml is more suitable.
Please don’t misunderstand me. As I tried and mentioned before – the whole disarmony here is more based on human and not on technical factors. I am on your side but for sure for different reasons. EL7 has changed to much not only the boot process. The cluster stack, IPsec implementation and httpd config syntax is different, NM for network, gnome-shell, missing TB and so on. Further investigation and adaptation of our deployment processes are in progress. So, EL6 will stay here as a primary plattform.
Peace.
I have. And I know how to. Ascii sort order is a straightforward concept for both humans and computers. I knew how to deal with scenarios with subtle issues like the network service ‘completing’
startup even though the connected switch was still doing spanning tree and the next network operation would fail. And now I don’t know how to tell how introducing some new arbitrary service will interact with an existing arbitrary set.
I have. And now those won’t work the same way. So why was using that abstraction good?
I read them at the important times. When things aren’t working. The ones shipped with the system have already had years of breakage gradually fixed in fedora. Our local apps will have to start from scratch.
You are just deferring the pain unless you plan to retire before EOL
for EL6 and foist the work off on someone else.
Not that simple. Smaller things none of which on its own make you make fundamental decision just accumulate, so after some critical mass you are there, the decision is made. I probably will not remember all of them, and definitely not in the order of their arrival.
1. SELinux (it doesn’t not pass _my_ “security has been enhanced”
estimate. It can be turned off on the fly, therefore it is as if it doesn’t exist. It adds tens of thousands of lines of code to the kernel, thus increasing chance of bugs and deteriorating security. And once my estimate was confirmed: “critical security update of SELinux”…)
2. The need to reboot linux box often (on average every 45 days in my observation…). I know, there are workarounds to patch on the fly…
3. Firewalld. I’m tempted to say: the philosophy it’s based on is flawed. Hey, this is our institutional internal network which is behind strict firewall, so we consider it “safe” zone… Nope. Wherever is the machine that _you_ do not administer, you better do not consider it “safe” or
“secure”. It uses the same iptables kernel module, right? And you can switch to using iptables instead if it, right? Only one day you end up running distro with stripped this, stripped that… Small thing in itself as I said.
4. Systemd. (“If it ain’t broke, don’t fix it”). I don’t want to start flames again…
And things add up gradually till you realize this. If I were willing to run M$ Windows I would pay for it, and will curse it legitimately as I
paid for it and my money entitles me to curse it. Now I run free open source system courtesy of uncounted number of generous developers. I
should be grateful to them and tolerate new routes they take – which is usually their only reward for their work… So, when I can not tolerate it any more (… accumulated above critical mass), I just go to another also open source system, thank goodness, they still do exist…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
In this phrase I was always wondering “as a kid” part is about “me” or about “my father” ;-)
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
And see this:
http://www.zdnet.com/lennart-poetterings-linus-torvalds-rant-7000034384/
Car vendors might behave differently however. Particularly if they obtained revenue from the necessary retraining and profited from the licensing procedures.
I’d rather have people look at this:
http://www.zdnet.com/linus-torvalds-and-others-on-linuxs-systemd-7000033847/
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
Ok I read the information. So as I understand it you are going have a computer that has multiple graphics cards with multiple keyboards and multiple mice divided into seats. Really?
Where do I buy this computer?
Yes, “much too cavalier about bugs and compatibility” is exactly right. Never mind the fact that you break everything you existing users have done, you might attract some people who like windows and don’t understand that it took decades of patches and reboots every Tuesday to reach any kind of stability with its monolithic approach.
Once upon a time, Steve Clark said:
Umm, Wal-Mart? Best Buy? You do realize that AMD/ATI Radeon chips (for example) in many systems can drive multiple display simultaneously? Add another USB keyboard/mouse/speakers, and you can set them up as multiple seats. Pretty much any modern PC can have more than one video card.
It is much simpler to run remote X sessions over a network for multiuser access and probably not much more expensive if you use older PCs as terminals. You do have to boot something, but x2go or freenx/NX are cross platform and have great remote performance. I’m surprised no one has made a mini-linux distro that boots straight to x2go for this purpose, but if they have, I haven’t found it.
Or a raspberry-pi for $50 if you don’t have an old PC.
Am 09.10.2014 um 16:57 schrieb Les Mikesell:
As I said, because you don’t known doesn’t mean its faulty -> human factor.
An interesting idea, that was born in your head, right?
I would suggest to keeping the facts on the table and concatenating opinions and other nonsense to device null.
“human factor” means it is extra work for someone who was already using a linux distribution because it did what they wanted – which doesn’t seem good to me either. And since I don’t understand it, I
don’t know how to prove that some arbitrary addition can’t cause a dependency loop when added to some arbitrary existing configuration. Do you?
No, it is a very real scenario for me. If I leave the upgrades for someone else it will probably be the end of linux use here.
The harsh reality of life is ‘real’ villains etc. don’t talk about harming someone else, they just do it or employ others to do it. Talking, discussions, proposals etc. are dangerous to the instigators because it creates implicating criminal evidence.
Systemd should be been much more widely discussed with the general Linux / Red Hat and clones community. I am sure beneficial improvements would have been proposed and, hopefully, implemented prior to Red Hat’s adoption.
That’s life. Few things are perfect :-)
Change is inevitable in life. Virtually everything changes including the eventual decline of our sun. Not sure about atomic weights or the value of Pi (3.142) or E=mc² Perhaps they break the rule that everything changes.
Being optimistic perhaps systemd will quickly change into something more welcoming to the vast majority of users.
—
Regards,
Paul. England, EU.
It is about fundamental approach. We always modularize things: split into smaller subunits each of the last doing its smaller task. This allows to make smaller things work reliably, and test these smaller things more comprehensively. As it is much smaller number of combinations of factors you need to repeat your test with in case of subunits. People use this approach for ages. Programs are split into subroutines. Rockets are built from to awful degree independent modules. We had this “modular” system V
boot until recently. We lost it. We got “iPhone, with whatever you can get in App store” instead. And not all of us are pleased by this change. And, BTW, there was one of the posts of MS Windows big fan on this list who welcomes this change; his post should have made everybody think…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
As a computer person of nearly 50 years experience (wish I was 20
again), I have long recognised the essential fact ….
* That if one gets it Right the first time, one inevitably saves an awful lot of time and an awful lot of effort subsequently trying to get it Right or even to “improve” the less than perfect item.
A little critical and constructive effort into the architecture and functionality of systemd would have brought considerable benefit. Perhaps the Fedora and RedHat persons might remember this advice when the next “wonder” is proposed. No reason not to want *genuine*
perfection for all.
Astonishing ! Its how I do my projects, programming and other work. Do you think our Fedora friends might adopt your revolutionary idea ?
Well, one suggestion is we set-up a working group to devise a replacement (or replacements) for systemd. The most difficult task is selecting a nice name for the systemd replacement.
No good anyone moaning when only action will bring change.
No. I think they see the huge market that Windows has and want to emulate their approach at the expense of anyone who liked unix and headless multi-user servers. But the people who like windows don’t really need a bad imitation either.
I’m curious on how you think that SysV init is modular in this approach and systemd isn’t?
A simple ‘ps uf -p 1’ on a couple of machines shows about 10x the resident memory use and 5x virtual on CentOS 7 vs. 5.x. And yet, the programs that it started don’t show any improvement for the extra cost.
So, the extra memory is being used by systemd processes?
What makes you think that systemd is not modular? Have you actually looked at its structure (let alone the code)? If you look inside /usr/lib/systemd/ do you see one big monolithic library which represents one big failure point, or do you see a few dozen dedicated small libraries, each doing one particular thing?
I don’t really see how systemd violates the “do one thing and do it well” philosophy. A lot of people seem to ignore the fact that systemd is *not* one binary executable which replaces init and tries to take control of everything, but rather a whole swarm of independent binaries, each in charge of one particular function of the OS. If one of them breaks for some reason, others will still keep functioning.
It appears to me that much of bashing of systemd is just FUD. One of the typical misconceptions is the disable vs. mask for services –
+1,, and whatever else is in vogue.
hear here! (or is that, here hear?)
Once upon a time, Marko Vojinovic said:
systemd (as PID 1) is not necessarily the problem. The problem IMHO is the systemd _project_ that appears to have a severe case of scope creep. They have swallowed up other projects (udev, dbus), reinvented wheels
(logging, ntp, network configuration), and appear to still be growing without bounds. When people don’t like some of the decisions of systemd developers, and the systemd project keeps taking over more of the core OS functionality, it is frustrating to watch.
I don’t quite disagree, but: a number of programs in Fedora/RHEL/CentOS
actually use (or possibly abuse) chkconfig to see if a service should be run from cron or by some other trigger. The systemd implementation is more clean about being used for “disable starting at boot”.
I agree that it ended up being different terminology than I think is most clear and ideal, but… c’mon, we’re sysadmins. If we start listing all of the things in Linux with that problem we will be here all day. Eventually, you just learn it, think, “whelp, another one for the that’s-a-little-weird file”, and move on.
There’s probably nothing wrong with these things in the context of a new/different OS oriented to servicing a single user who is assumed to own the world as a side effect of logging into a magical console device and starting a GUI where things can pop into existence and chat with other things easily. But it doesn’t have much to do with unix-like concepts.
What happens to ownership of a DVD or audio device when a different user logs in at the console – even if some other remote user wants to access them? The magic is more about ConsoleKit and PolicyKit than specifically systemd but the underlying dbus gunk seems to be merging together.
And the monthly reminder:
Please watch your replies to this list. Reindl is _not_ a subscriber and all you do is inject his diatribes back here where he’s not wanted thus circumventing his removal.
Thank you.
John
You think that nobody on that project thought about this before?
It’s not just remote X sessions. You want at least USB and audio redirection and also a decent 3D performance.
We currently do that using spice for VMs, I don’t know how feasible it is to run it on a real hardware.
There are some good pro’s on this setup:
– this installation is physically simpler than having 4 full computers as it requires 1/4 of the wall plugs and network points
– no single point of failure (as in: 4 seats down is okay), if you compare with ones using x2go and similar (application server)
– easily scalable: need more seats? buy 1 computer more, you have +4
seats, and you’re good. No server needs to be re-evaluated.
– easier to maintain, as you maintain 1/4 of the systems you would otherwise.
– very cost effective with commodity hardware, that everyone knows how to deal with.
– vendor independent
And probably many others that I forgot :)
Not saying it’s the best, though. Just saying that yes this is a good project that is well plotted and has its audience.
Marcelo
Yes but you have to be physically close to the main cpu. What about distractions from other people sitting right next to you?
Playing music, etc.
On 10/13/2014 7:17 AM, Steve Clark wrote:> Yes but you have to be physically close to the main cpu. What about
That’s not all that different from modern cube farms. You learn to tolerate or ignore other people, or more ideally, collaborate well with your closest co-workers.
Where I work, there are people sitting side-by-side at folding tables (business has picked up faster than physical facilities can keep up with). In our case, they’re all using zero clients and virtual desktops. However, it’s exactly the kind of setup where a multi-seat computer might make sense for other companies or schools that don’t have our virtualization expertise.
Being able to grab your existing desktop remotely with all open windows and long-running programs intact is a big plus, though – and you get that for free with NX or x2go. Can you connect remotely to your VM host some other way if you aren’t at the special desktop?
How much different is that from VNC? Just curious.
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
VNC just does a basic bit-copy of the screen for the remote side so it is not particularly efficient, although it might be better if you are mirroring a real console instead of running a virtual session set up by VNCserver. NX/x2go have a full proxy/stub X server and client at both ends with tunable caching/compressiion on the remote side.so things like font rendering and block moves for scrolling are much faster. X2go also can map audio/drives/printers if you want.
Both are packaged and fairly easy to try on CentOS 6. On 7, only x2go is available and it has a problem with the 3d requirement of Gnome3 so you have to use KDE or install MATE from EPEL.
I didn’t mean to imply that you use a multi-seat computer to get to a desktop served by a remote machine, though you could certainly do that if you wanted. Everyone still needs a machine to function as an endpoint for the remote desktop to be delivered to, though. You use a multi-seat computer when you don’t have enough computers to give everyone their own machine.
Like William Gibson said, “The future is already here – it’s just not evenly distributed.” I have enough computers that I could make furniture out of them, but I’m sure there’s some cash-strapped school district using donated hardware that would jump at the chance to have their computer lab serve ten students at a time instead of five.
— -Chris
I haven’t used x2go, but NX is way faster than VNC over a slow link
(e.g., home DSL, hotel wifi) in my experience.
–keith
Ah, just remembered one thing. If memory serves, FreeNX dates back to the same time that multiseat project was started. They were developed targeting a very similar use case, but different in the end, while multiseat had a faster time-to-market, I think.
Places like some public libraries, like we have here in Brazil, uses multiseat. Keeping these running is much easier than having someone with the knowledge to fix a virtualization host when it goes rogue. These places here have 0 tech people working in there..
Note that multiseat doesn’t even require a network at all. You can always open up a text editor and do some writing, or some spreadsheet calcs…
It’s been a while since I last used FreeNX. I didn’t recall it being able to map drivers back then but good to know that x2go can do it, thanks. Back then it already was indeed much faster than VNC and/or remote X.
Marcelo
The K12LTSP project might have been a good fit several years ago. This was a respin of fedora or CentOS 4/5 distributions that would come ready to PXE-boot diskless PCs as thin clients and host their sessions. I think you could make them auto-login to a kisok type application if you didn’t want individual user logins. That involved more hardware, of course, but a lot of places used old donated boxes with the disks removed and only needed one real server per classroom of 30 or so. The project evolved (or devolved, depending on how you look at it) into packages that you have to install on a stock system and that now have more of a fat-client with local apps approach
(k12osn).
Sure, but networks are cheap and reliable and easy to run longer distances than vga cables.
On a one-hop network, even straight remote X is reasonable for thin-type clients. The main problem remaining is doing full resolution streaming video to many desktops from a multiuser host –
which the multi-video card approach should do for some number of seats.
But, I’m kind of surprised that someone hasn’t done a raspberry-pi type device that boots directly into x2go and comes out cheaper than a video card per seat. Haven’t needed one badly enough to build it myself yet.
It should be trivial to set up an actual RPi to do that.
One of the options during installation is to boot into text mode instead of graphical mode. Once it
Not really. The beauty of the original K12LTSP respin was that just you did a normal fill-in-the-form install pretty much like any fedora/CentOS in a server with 2 NICs and you could plug in some diskless PCs and they came up working with applications ready to go. Compared to that, there’s still a lot of do-it-yourself assembly required. I recall from the early K12LTSP mail list days that quite a few teachers were able to set up a classroom with very little admin knowledge.
Agreed on the hardware front, but couldn’t this be a canned image you copy to an SD card with some way to edit the target IP address in place without needing to rebuild it?
—
Les Mikesell
lesmikesell@gmail.com
If you insist on having a whole OS dedicated to this, I guess you could go fork Raspbian (http://www.raspbian.org/) and add this stuff to the installer.
Compare that amount of work to:
1. Install Raspbian from NOOBS
2. Change boot option so it doesn
I’m not insisting on anything, I just don’t see a typical classroom teacher doing that – in addition to knowing what to set up on the server side to match. Where they could just follow some simple instructions with the LTSP install and come up working.
Agreed, but K12LTSP worked mostly as-is for a lot of districts.
I don’t need it myself – I’m just surprised that someone hasn’t done it already in a way that would be useful without everyone having to start from scratch themselves. And maybe that’s not even the best approach – maybe a ‘fat’ client with NFS-mounted or cloud-like storage would be better if there is an automated way to keep it up to date without letting the user modify things. I guess these days you’d have to balance the cost of administering the things against buying a chromebook.
—
Les Mikesell
lesmikesell@gmail.com