Systemd provides a replacement for things like ConsoleKit. It further makes various things much simpler for GNOME. It provides dbus interfaces. We're not forcing, it is that we do like to share the amount of work a slight bit. Meaning: those "portability" requests are nice, but almost everyone uses Linux. Isn't it logical that after 10+ years of making things also possible on BSD, we now have Linux as main focus?
All those portability claims and that there somehow good: you're just doing things for lowest common denominator. Requiring a huge amount of effort to change or develop low level bits across BSD and Linux.
Now systemd comes along and does a lot of things for us. While you'd still have dbus. We now have people who solely developed for Linux and they're installing FreeBSD, that's indicative on how much development resources there are on BSD!
You imply that GNOME developers (none running or knowing anything other than Linux) are bad because they're not also developers for BSD. I've seen other people claim that us focussing mainly on Linux will somehow reduce the amount of help we get from BSD and so on. Yet practically speaking, the amount of developer resource we received during most of GNOME 2.x from BSD is pretty low (mainly implementations; hal, upower backends, etc). Plus often lagged behind a bit.
Recently we're making clear that we focus on Linux. We're still as open to portability as before, but someone really has to put in an effort to do this. Putting that with people who have only used Linux and mainly care for Linux is a bit strange IMO, especially after all those years.
Given that *BSD often did lag behind and did eventually provide implementations, the situation has not changed much. Expect for the often repeated claim of "forcing things".
This incoherent rant as a response to a single sentence isn't really helping your case.
"All those portability claims and that there somehow good: you're just doing things for lowest common denominator. Requiring a huge amount of effort to change or develop low level bits across BSD and Linux."
I gave an explanation on how things are, it was not a rant. I don't understand how you could see it as a rant.
Regarding your question: GNOME does a lot of infrastructural work (freedesktop.org, and other things). A lot of the times things are just not available. Thing e.g. of colord (colour management), PackageKit, hal (dead :-P), ConsoleKit (dead :-P), etc. Making that work on Linux is one thing (already loads of differences). Making that work for others as well, while sometimes there is not much available only increases the time you have to spend on things.
not really he's implyming gnome developers are bad, because they shut off any kind of dialog. it's kind of the same reason why gentoo folks developed eudev.
who cares about bsd? you can't even use different distros anymore, because these folks are destroying the unix philosophy. wanna run gnome? sure, replace everything with systemd.
EDIT: Firstly this post was more about "the GNOME attitude" than systemd. However, I've already read the biggest myth post, so please stop responding with that post. So far all our software engineering efforts have lead to the conclusion that whenever a system leads to too much complexity it because cumbersome to use extend, and prone to errors. i would argue that you can build a single piece of software that is modular without splitting it into multiple parts, but systemd isn't that.
In fact, a lot more companies are moving away from big monolithic blocks. No one ever said that single components should be built in a non-connected way. in fact, composability is a big part of good components. where ever did you get the idea that its the opposite?
why do you think imlib2 got so successful? but hey, care to explain to me what the advantage of merging udev with systemd was?
Not saying I disagree with you, but you mention the "unix philosophy".
These last few years I've seen a lot of people repeat that sentence over and over again, but I ask you, what is the "unix philosophy"? Treating everything as a file? Do one thing and do it well? Develop beautiful software, not bloated and simple?
Every single person has a different idea of what this mythical "unix philosophy" actually is.
Certainly, all these practices apply well on servers but the world is moving forward and we have much much more complex devices that aren't just mainframes with terminals anymore.
We are not dealing with software anymore, we are dealing with systems. And we want performance and stability. Developing single components in a totally individual and non-connected way, oblivious of their environment does not always work anymore.
Sometimes you need to build an environment, not just a single component. You need to have software that interfaces and integrates properly.
Does the Unix philosophy go against that? I don't know. The point is that you should always question things, not just repeat "but the Unix philosophy says...", because systems evolve.
It's also an excuse to avoid good design. It requires more effort to design a system that doesn't have a strong coupling between the components and a clear set of abstractions.
Does it mean that systemd is bloated ? I don't know, I would have to study the problem space closer.
Even if we can't define the "UNIX philosophy" exactly, I think we can still look at a given software system and see whether or not it generally approaches that set of ideals as a whole.
When compared to UNIX-oriented software of the past several decades, I think it's clear that modern GNOME is very different. Systemd is also very different. Being so different in so many core ways means that they inherently don't adhere to the practices and ideals we can consider to be typical of software adopting the "UNIX philosophy".
I think you skipped over the bit where I explained things.
E.g. you're saying "rely on systemd", while I already mentioned we rely on dbus interfaces. However, there is only sort of 1 implementation (other is a fork), so nothing for *BSD.
Now eudev as explained during FOSDEM of 2013 was only started because they didn't understand plus they wanted to understand C. There hasn't been any progress. Further, the packaging of udev in Gentoo caused some bad assumption.
Merging of udev and systemd had loads of advantages, but if you bring up eudev again, I'm guessing you're missing out on 12 months of explanations (I attended the eudev presentation at IIRC FOSDEM 2013). Greg (the kernel dev) mentioned during FOSDEM 2014 that he still follows eudev development and there is not much going on. Sometimes bugfixes for bugs introduced by eudev.
Regarding most of your other replies:
Please explain properly what you mean. "Unix Philosophy" could mean anything for you. I'm more than willing to discuss, but that's impossible if I don't understand you.
> Now eudev as explained during FOSDEM of 2013 was only started because they didn't understand plus they wanted to understand C. There hasn't been any progress. Further, the packaging of udev in Gentoo caused some bad assumption.
> but if you bring up eudev again, I'm guessing you're missing out on 12 months of explanations (I attended the eudev presentation at IIRC FOSDEM 2013). Greg (the kernel dev) mentioned during FOSDEM 2014 that he still follows eudev development and there is not much going on. Sometimes bugfixes for bugs introduced by eudev.
not surprised, i wouldn't be surprised if the project was dead already. i said the reason it was forked not the reason they maintain it or did i?
> Regarding most of your other replies: Please explain properly what you mean. "Unix Philosophy" could mean anything for you. I'm more than willing to discuss, but that's impossible if I don't understand you.
sorry, isn't that written in my post? also, sorry, since everyone seems to be sensible to the words unix philosophy even though we're still chaining, and composing those very utilities to this date. forget that word.
make it good design. modular, composable, well abstracted(but not too abstracted software).
EDIT: also, sorry, i didn't mean to start a rant page on systemd here. i use it on my archlinux, and i'm quite happy there. but I stopped using gnome a long time ago. not only that, i also stopped using gnome components, even though for a long time i was using gnome-session with awesome after going down the tiling wm route.
I get the impression that you think I am trying to argue, if so: it is not the case, I'm trying to understand and at the same time to explain. I often explain the same thing various times because I follow lots of mailing list and try to give explanations where I can. So in case my answer is sometimes long is not regarding starting an argument, it is because I gave similar explanations various times before and I want to be thorough.
I've followed eudev quite a bit. The initial email and following discussion was very confusing. The LWN link you give is based on various misunderstandings. Eventually 3 eudev developers gave the reasoning I gave before. That's why I find it quite interesting.
Those misunderstandings and assumptions is is also why I try to explain things like "portability" as good as possible: To avoid anyone getting bad impressions. E.g. if you'd just read any news site casually you'd assume we're totally pushing systemd, while it is far more nuanced.
EDIT: funny, i never saw lu_zero in person. not quite what i expected.
but anyway, it looks to me like they clearly did target some systemd attributed issues. they were being very careful not to bash lennart and his friends though, it almost looks like the systemd people are bashing them more than the other way round.
This is FUD, and the worst is that you don't even know it. Geolocation is a NEW feature in GNOME 3.12, so you don't have to wait a release to disable Geolocation, because you don't have it yet. The article point to a commit during the development process.
Actually, almost everyone uses Windows. Of those that do not, majority uses OSX.
As someone who's been a Linux desktop user since the early 2000s and was laughed out loud because everyone uses Windows so why even bother, I'm saddened by seeing the same logic from the GNOME team.
(PS. Obvoisly we're talking about desktop here so Android doesn't count).
I actually meant the people helping out at GNOME. So the developers, translators, etc. If everyone helping out in GNOME, then I find it logical that the focus is mainly on Linux.
I .. don't even understand how you can possible misunderstand/misrepresent that statement.
Don't ignore the context: Gnome developers run, according to bkor (and I'd say he should know the numbers..), more or less completely on Linux. That's what they work with, what they know. THAT is what 'almost everyone uses Linux' means.
Your random 'Hey, what about Windows?' statement is absolutely irrelevant, period.
>As someone who's been a Linux desktop user since the early 2000s and was laughed out loud because everyone uses Windows so why even bother, I'm saddened by seeing the same logic from the GNOME team.
Well, below some number of users (userbase) the logic becomes solid.
I'm sure you'd agree too.
E.g Would you go out of your way to cater for an OS with 10 users? At some point there are diminishing returns. Is that 100? 1000? 10000? What NetBSD has? I don't know.
But just because you found it unfair to be laughed out because "everybody used Windows", doesn't mean it's not pragmatic to stop catering to niche OSes at some point, if your main focus is say Linux and getting compatible with them eats on your time and effort.
>Isn't it logical that after 10+ years of making things also possible on BSD, we now have Linux as main focus?
Sure. But then don't keep lying and pretending it is portable. If you are writing linux only software then don't get mad when BSD people call it linux only software.
Did you actually read the blog post? It seems like you didn't. Did you also read https://wiki.gnome.org/ReleasePlanning/WhatWeRelease? Do you understand what I mean with main focus? It means we try to ensure that it is portable, but we don't go out of our way to make it already works anywhere other than Linux. While at the same time, some people are actually doing that more than ever before.
Your summary of "keep lying" is so far off that I guess you just didn't read the blogpost, nor my comment in any detail.
I am responding to what you just said, not the blog post. Again, feel free to write linux only software. But if you do that and then call BSD people liars when they call your linux only software linux only software, then expect to be called on it.
I'd say you're right, but it seems bkor (and I feel somewhat the same) read some judgement in your summary.
My summary would be
So basically "We're really focusing on Linux officially now, since that's what we're accustomed with. But look at these great ports to two different BSDs!"
Paired with the blog post at [1] (linked elsewhere in this thread) the situation doesn't look as ugly as people are trying to imply.
An OS which doesn't have systemd-like functionality probably isn't very modern and is thus not an interesting porting target for GNOME. I expect the BSDs will get such functionality sometime in the near future.
You suspect wrong. Being new does not mean the same as being modern. Systemd solves a problem that doesn't exist. You are never going to see that kind of nonsense in openbsd for example.
Not entirely related – but god how bad Gnome 3 still looks. I was under the impression that they would make some changes to the initial UI design. I'm sure they have, but it still looks terrible. I've not found a lot of UI design that hurts my eyes so much. I always get stuck on the upper left corner and the edge curvature of the menubar at the corner of the screen (activities: http://blogs.gnome.org/mclasen/files/2014/02/gnome-on-freebs... ). I feel a bit bad when hating on open-source projects, but they really need to hear this.
I agree - I was only willing to switch to Linux full time once Gnome 3 was released. I find Unity garish and Gnome 2 and KDE to be far too tweak happy with lots of ugly rough edges.
It looks ok... much better than the hideous default KDE in my opinion. The fact that it doesn't have the functionality of even Windows NT 4.0 (after years of pruning) is the bigger problem imho.
What a remarkably dishonest argument. The example of gnome running on openbsd is precisely because the openbsd ports people patched the hell out of it to remove all the non-portable crap. Leaving a gnome with a bunch of features gone. And notice Antoine who did that work and provided that screenshot is one of the people who thinks there is no future for gnome on non-linux systems: https://lists.debian.org/debian-devel/2013/10/msg00580.html
Why do you present a false dichotomy? The gnome team (and presumably at least some gnome users) consider them features. Those features are implemented as non-portable crap. Saying gnome works fine on openbsd and pointing to part of gnome working only after an openbsd guy ripped out big chunks of it is just silly.
In the blogpost referenced here as well as another blogpost, OpenBSD and FreeBSD developers appreciate the work that Ryan has been doing. Calling everything "crap" is a bit easy. What I did learn (since today) is that we were waiting for them (*BSD) to contact us (GNOME), while once only 1 person contacted them, we got a lot of cooperation going.
Meh. I understand the BSDs are great for server applications and that's where I'd happily use them.
But the market for a free desktop is tiny--for a BSD desktop (!OSX) miniscule. I see no problem with letting the burden of maintaining ports fall on those who would use them. (Assuming what he wrote was true, that they aren't making ports to *BSD difficult.)
42 comments
[ 4.4 ms ] story [ 98.8 ms ] threadAll those portability claims and that there somehow good: you're just doing things for lowest common denominator. Requiring a huge amount of effort to change or develop low level bits across BSD and Linux.
Now systemd comes along and does a lot of things for us. While you'd still have dbus. We now have people who solely developed for Linux and they're installing FreeBSD, that's indicative on how much development resources there are on BSD!
You imply that GNOME developers (none running or knowing anything other than Linux) are bad because they're not also developers for BSD. I've seen other people claim that us focussing mainly on Linux will somehow reduce the amount of help we get from BSD and so on. Yet practically speaking, the amount of developer resource we received during most of GNOME 2.x from BSD is pretty low (mainly implementations; hal, upower backends, etc). Plus often lagged behind a bit.
Recently we're making clear that we focus on Linux. We're still as open to portability as before, but someone really has to put in an effort to do this. Putting that with people who have only used Linux and mainly care for Linux is a bit strange IMO, especially after all those years.
Given that *BSD often did lag behind and did eventually provide implementations, the situation has not changed much. Expect for the often repeated claim of "forcing things".
"All those portability claims and that there somehow good: you're just doing things for lowest common denominator. Requiring a huge amount of effort to change or develop low level bits across BSD and Linux."
I don't even know what you're trying to say here.
Regarding your question: GNOME does a lot of infrastructural work (freedesktop.org, and other things). A lot of the times things are just not available. Thing e.g. of colord (colour management), PackageKit, hal (dead :-P), ConsoleKit (dead :-P), etc. Making that work on Linux is one thing (already loads of differences). Making that work for others as well, while sometimes there is not much available only increases the time you have to spend on things.
who cares about bsd? you can't even use different distros anymore, because these folks are destroying the unix philosophy. wanna run gnome? sure, replace everything with systemd.
wanna disable geolocation in gnome? no problem just wait for the next release http://www.phoronix.com/scan.php?page=news_item&px=MTYwNTM
EDIT: Firstly this post was more about "the GNOME attitude" than systemd. However, I've already read the biggest myth post, so please stop responding with that post. So far all our software engineering efforts have lead to the conclusion that whenever a system leads to too much complexity it because cumbersome to use extend, and prone to errors. i would argue that you can build a single piece of software that is modular without splitting it into multiple parts, but systemd isn't that.
In fact, a lot more companies are moving away from big monolithic blocks. No one ever said that single components should be built in a non-connected way. in fact, composability is a big part of good components. where ever did you get the idea that its the opposite?
why do you think imlib2 got so successful? but hey, care to explain to me what the advantage of merging udev with systemd was?
These last few years I've seen a lot of people repeat that sentence over and over again, but I ask you, what is the "unix philosophy"? Treating everything as a file? Do one thing and do it well? Develop beautiful software, not bloated and simple? Every single person has a different idea of what this mythical "unix philosophy" actually is. Certainly, all these practices apply well on servers but the world is moving forward and we have much much more complex devices that aren't just mainframes with terminals anymore. We are not dealing with software anymore, we are dealing with systems. And we want performance and stability. Developing single components in a totally individual and non-connected way, oblivious of their environment does not always work anymore. Sometimes you need to build an environment, not just a single component. You need to have software that interfaces and integrates properly.
Does the Unix philosophy go against that? I don't know. The point is that you should always question things, not just repeat "but the Unix philosophy says...", because systems evolve.
Furthermore, I'll point towards this page http://0pointer.de/blog/projects/the-biggest-myths.html
It might be relevant, it might be not, but it's a good read nonetheless.
Does it mean that systemd is bloated ? I don't know, I would have to study the problem space closer.
When compared to UNIX-oriented software of the past several decades, I think it's clear that modern GNOME is very different. Systemd is also very different. Being so different in so many core ways means that they inherently don't adhere to the practices and ideals we can consider to be typical of software adopting the "UNIX philosophy".
E.g. you're saying "rely on systemd", while I already mentioned we rely on dbus interfaces. However, there is only sort of 1 implementation (other is a fork), so nothing for *BSD.
Now eudev as explained during FOSDEM of 2013 was only started because they didn't understand plus they wanted to understand C. There hasn't been any progress. Further, the packaging of udev in Gentoo caused some bad assumption.
Merging of udev and systemd had loads of advantages, but if you bring up eudev again, I'm guessing you're missing out on 12 months of explanations (I attended the eudev presentation at IIRC FOSDEM 2013). Greg (the kernel dev) mentioned during FOSDEM 2014 that he still follows eudev development and there is not much going on. Sometimes bugfixes for bugs introduced by eudev.
Regarding most of your other replies: Please explain properly what you mean. "Unix Philosophy" could mean anything for you. I'm more than willing to discuss, but that's impossible if I don't understand you.
shrug, maybe, links? https://lwn.net/Articles/529314/
> Merging of udev and systemd had loads of advantages,
does it? http://www.mail-archive.com/systemd-devel@lists.freedesktop....
> but if you bring up eudev again, I'm guessing you're missing out on 12 months of explanations (I attended the eudev presentation at IIRC FOSDEM 2013). Greg (the kernel dev) mentioned during FOSDEM 2014 that he still follows eudev development and there is not much going on. Sometimes bugfixes for bugs introduced by eudev.
not surprised, i wouldn't be surprised if the project was dead already. i said the reason it was forked not the reason they maintain it or did i?
> Regarding most of your other replies: Please explain properly what you mean. "Unix Philosophy" could mean anything for you. I'm more than willing to discuss, but that's impossible if I don't understand you.
sorry, isn't that written in my post? also, sorry, since everyone seems to be sensible to the words unix philosophy even though we're still chaining, and composing those very utilities to this date. forget that word.
make it good design. modular, composable, well abstracted(but not too abstracted software).
EDIT: also, sorry, i didn't mean to start a rant page on systemd here. i use it on my archlinux, and i'm quite happy there. but I stopped using gnome a long time ago. not only that, i also stopped using gnome components, even though for a long time i was using gnome-session with awesome after going down the tiling wm route.
I've followed eudev quite a bit. The initial email and following discussion was very confusing. The LWN link you give is based on various misunderstandings. Eventually 3 eudev developers gave the reasoning I gave before. That's why I find it quite interesting.
Those misunderstandings and assumptions is is also why I try to explain things like "portability" as good as possible: To avoid anyone getting bad impressions. E.g. if you'd just read any news site casually you'd assume we're totally pushing systemd, while it is far more nuanced.
Also, I am not the author of the blog / blogpost!
EDIT: funny, i never saw lu_zero in person. not quite what i expected.
but anyway, it looks to me like they clearly did target some systemd attributed issues. they were being very careful not to bash lennart and his friends though, it almost looks like the systemd people are bashing them more than the other way round.
This is FUD, and the worst is that you don't even know it. Geolocation is a NEW feature in GNOME 3.12, so you don't have to wait a release to disable Geolocation, because you don't have it yet. The article point to a commit during the development process.
Actually, almost everyone uses Windows. Of those that do not, majority uses OSX.
As someone who's been a Linux desktop user since the early 2000s and was laughed out loud because everyone uses Windows so why even bother, I'm saddened by seeing the same logic from the GNOME team.
(PS. Obvoisly we're talking about desktop here so Android doesn't count).
Don't ignore the context: Gnome developers run, according to bkor (and I'd say he should know the numbers..), more or less completely on Linux. That's what they work with, what they know. THAT is what 'almost everyone uses Linux' means.
Your random 'Hey, what about Windows?' statement is absolutely irrelevant, period.
Well, below some number of users (userbase) the logic becomes solid.
I'm sure you'd agree too.
E.g Would you go out of your way to cater for an OS with 10 users? At some point there are diminishing returns. Is that 100? 1000? 10000? What NetBSD has? I don't know.
But just because you found it unfair to be laughed out because "everybody used Windows", doesn't mean it's not pragmatic to stop catering to niche OSes at some point, if your main focus is say Linux and getting compatible with them eats on your time and effort.
Sure. But then don't keep lying and pretending it is portable. If you are writing linux only software then don't get mad when BSD people call it linux only software.
Your summary of "keep lying" is so far off that I guess you just didn't read the blogpost, nor my comment in any detail.
My summary would be
So basically "We're really focusing on Linux officially now, since that's what we're accustomed with. But look at these great ports to two different BSDs!"
Paired with the blog post at [1] (linked elsewhere in this thread) the situation doesn't look as ugly as people are trying to imply.
1: http://undeadly.org/cgi?action=article&sid=20140219085851
I'm so glad there's XFCE.
http://www.micahcarrick.com/images/migrating-to-gnome-3/tint...
It always felt weird to me too, but well.. subjectivity.
Make up your mind. Was what they removed "crap" or was it "a bunch of features"?
Really? All the patches that were sent upstream and ignored for years wasn't contact?
But the market for a free desktop is tiny--for a BSD desktop (!OSX) miniscule. I see no problem with letting the burden of maintaining ports fall on those who would use them. (Assuming what he wrote was true, that they aren't making ports to *BSD difficult.)