|
|
Subscribe / Log in / New account

Dirk Hohndel at Akademy (KDE.News)

KDE.News sat down for an interview with Dirk Hohndel, Intel's chief Linux and open source technologist, at Akademy. Hohndel talked about various things including Aaron Seigo's keynote, happiness at work, and his relationship to KDE. "Dirk H: I know a lot of the early KDE community members like Kalle and Matthias. I used KDE software from the start to the 2.0 release, and lost contact for a while. I'm interested in the Linux client so since a couple of years I'm trying to figure out where it is going. I follow both Gnome and KDE development, see what is going on in the open source world. I've been repeatedly trying the latest KDE Plasma Desktop (which still fails for me for a number of reasons). The KDE community interestingly has undergone a lot of changes - very much unlike the Linux kernel where the 'oldtimers' are still very much involved. Seriously, of the 20 initial Linux developers, 15 are still very active."

to post comments

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 20, 2010 19:13 UTC (Tue) by sdalley (subscriber, #18550) [Link] (44 responses)

Dirk H: Well, all projects have their ups and downs and KDE is in a very good place. I loved the keynote, where Aaron said we should focus on user experience, elegance, consistency. What is stopping me to use it is as much the little details as it is habits I have - it is not the technology per se. And to be fair and honest, whenever I switch to a KDE based set of apps I do it on Fedora. I have customized my Gnome desktop for the last 6 years and in the Plasma Desktop in Fedora I get the default configuration. So the hard part is that many things I expect to work just don't. I haven't figured out how to use a hotkey to suspend my machine, for example. You have to go through menus. There is the lock & shutdown widget but it doesn't have a suspend button - what were you guys smoking while you wrote it? Or take KNetworkManager, its integration is weird. And if I have a panel scaled to have two rows of buttons for my apps, the other icons and the clock get HUGE. And there was some more stuff driving me nuts, I don't remember. After a week I still hadn't gotten things to work as I liked it so I gave up. I'll try it again, but for now - not my cup of tea.
Thanks for that heads-up Dirk. You've saved me the bother of trying KDE. I'll look again in a year or so.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 20, 2010 19:30 UTC (Tue) by dlang (guest, #313) [Link] (25 responses)

a year from now you will be able to find another post from someone who didn't like KDE and/or couldn't figure out how to tweak it. (for example, keybindings are under system settings, keyboard and mouse for things like setting a key to trigger suspend. not something I've bothered with, but I don't see why it would be so hard to find)

you can find similar posts from people about any desktop (including microsoft and OS/X), especially from people who have used a different system for many years and tweaked it extensivly. The problem is usually less that the new system can't be tweaked, and more in that it's just different from their old system.

what part of his post convinced you that KDE isn't worth trying?

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 20, 2010 21:36 UTC (Tue) by xxiao (guest, #9631) [Link] (1 responses)

it seems redhat/ubuntu/debian are all using gnome by default, that also puts kde to a secondary spot. at one time gnome seems too basic to me, but after switched to it a few years ago(i.e. after a few kunbutus I finally switched to the default ubuntu, esp the LTS release), I have no reason to jump back, gnome does the job well so far, easy to use, powerful enough for daily development.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 20, 2010 21:52 UTC (Tue) by dlang (guest, #313) [Link]

nothing at all wrong with that. the post I replied to seemed to be indicating that the user wouldn't even look at KDE due to the issues the poster before him had with KDE.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 9:43 UTC (Wed) by jwakely (subscriber, #60262) [Link] (10 responses)

> especially from people who have used a different system for many years and tweaked it extensivly

What's annoying is when you've used the *same* system (KDE) for more than a decade, and it changes things for no discernible reason. KDE changes something with every release. The latest thing to stop working for me is NetworkManager, I think I'm supposed to use KNetworkmanager now but that's vastly inferior, as mentioned in the article.

KDE has lost its way and doesn't provide the value it used to, but won't admit it.

From the article: "If you keep breaking things you start to lose people at some point!"

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 12:40 UTC (Wed) by mpr22 (subscriber, #60784) [Link] (2 responses)

Out of curiosity, what was the default action for Ctrl-Esc in KDE3?

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 13:31 UTC (Wed) by nye (subscriber, #51576) [Link] (1 responses)

I believe by default it opened some sort of system/process monitoring application.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 14:31 UTC (Wed) by halla (subscriber, #14185) [Link]

It still does.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 13:43 UTC (Thu) by jospoortvliet (guest, #33164) [Link] (6 responses)

I think you must've misunderstood something.

NetworkManager is (besides being a horrible piece of shit) the underlying infrastructure used by most distributions for wireless network management. It was developed by a Gnome dev, so he also called the GUI "NetworkManager" because Gnome is, after all, the only linux desktop provider.

Then for KDE KNetworkManager was developed. Now it has been integrated in Plasma, and as soon as KDE has released Plasma 4.5 you'll find it there, and KNetworkManager as separate applet deprecated.

If you don't run Plasma 4.5 I'd recommend to use the gnome NetworkManager, it works as good as is to be expected. Other wise use the plasma applet. And KNetworkManager (while it works in most cases) hasn't been maintained for a while now - so it's probably not your best bet.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 21:07 UTC (Thu) by nix (subscriber, #2304) [Link] (4 responses)

And if your machine isn't a laptop or mobile device, run screaming from NetworkManager and all its ilk. I've resorted to pinning the damn thing at priority -1 on every one of my machines (or analogues on non-Debian-derived distros) because otherwise it gets pulled in by upgrades and then on the next boot my network doesn't work because NetworkManager has jumped in and blown away my static routes and customized MTUs and traffic shaping and everything else it could get its mucky little hands on.

IMNSHO it should be something that is installed by default only on laptop-like spins of distros, and is otherwise kept well out of the way. It's downright toxic to system stability on desktops and servers.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 21:17 UTC (Thu) by dlang (guest, #313) [Link]

hear hear.

even on my laptop I kill it off (with the exception that while I am on the road I may turn it on for a bit)

on my thinkpad it's too eager to decide something is wrong with the wireless card and quit working, while it works just fine if I hard-code the parameters into /etc/network/interfaces.

even with my wired network, both networkmanager and wicd turn a 'network cable was unplugged for a few seconds' into 'shutdown the network, for the user to do a manual disconnect/reconnect before anything will work again' situation.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 23, 2010 11:29 UTC (Fri) by sorpigal (guest, #36106) [Link] (2 responses)

Why would you foist crap like networkmanager on laptop users? If it's junk (and it is) then it's junk. I use wpa_supplicant directly on my laptop and have no problems with it. If you splashed a decent UI on top of that then you'd have wireless issues taken care of.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 29, 2010 17:28 UTC (Thu) by Epicanis (guest, #62805) [Link] (1 responses)

"If you splashed a decent UI on top of [wpa_supplicant] then you'd have wireless issues taken care of."

I've been happily using WICD for my network handling. It feels like a simpler, "cleaner" solution and has worked great for me for the last couple of years. Plus, if I want my computer to remain connected to the network when I log off, now it does...

It seems to operate great with KDE, too.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 30, 2010 7:57 UTC (Fri) by jospoortvliet (guest, #33164) [Link]

yes but it doesn't support a lot of things, like 3G, modems and stuff.

I hate NM but there isn't really a much better solution (for now). And there now is a NM commandline control thing, if only it kept the network up when there was a Xorg crash...

Dirk Hohndel at Akademy (KDE.News)

Posted Sep 2, 2010 18:25 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link]

Aside from being a very trollish comment, this is also incorrect. The GUI is called nm-applet. NetworkManager is just the daemon. KDE UI's use NetworkManager as well underneath.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 10:20 UTC (Wed) by sdalley (subscriber, #18550) [Link] (10 responses)

what xxiao said. KDE is great I'm sure for hackers and visionaries. It's not any one thing Dirk said, just the strong overall impression that there is still a lot of churn in the UI details, not to mention the infrastructure stuff like nepomuk that is still bedding in. OK, the blemishes are small individually, but they add up to death by a thousand cuts. Ubuntu Lucid/gnome benefits from their "thousand paper cuts" fixes, and this raises the bar considerably for KDE. I guess in my middle age I just want to get on with the job more rather than be always tripping over small stuff. Maybe KDE will one day take a break from being all visionary and do a 10000-paper-cuts exercise.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 13:37 UTC (Wed) by nye (subscriber, #51576) [Link] (9 responses)

>OK, the blemishes are small individually, but they add up to death by a thousand cuts. Ubuntu Lucid/gnome benefits from their "thousand paper cuts" fixes, and this raises the bar considerably for KDE.

Indeed. At one point I decided to sit down for a day and document all the little things that cut me, to report them. I was pretty much laughed out of the room, so to speak - 'stop wasting time with pointless little things like that'. It all adds up to a desktop that's basically unusable.

(Disclaimer: I haven't tried KDE4 in about a year, since I decided I would give it *at least* two more years to reach a stage I would be happy to call 'beta'. Possibly there has been real improvement already, but I'm not optimistic enough to check again yet.)

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 14:36 UTC (Wed) by anselm (subscriber, #2796) [Link] (8 responses)

I was pretty much laughed out of the room, so to speak - 'stop wasting time with pointless little things like that'.

Problems with KDE are either small, in which case fixing them would be pointless and trivial, so nobody seems to bother (after all there are bigger fish to fry), or they are big, in which case nobody wants to tackle them and everybody waits around for someone to rewrite the whole subsystem in question from scratch, since the original authors have lost interest long ago and the existing code is too crufty for others to work with (it's likely to be C++, so what do you expect?). The reimplementation will take years, and the result will be software that is sort of like what was there to begin with, but replaces crucial functionality of the previous incarnation by cool-sounding new abstraction layers that don't quite do what they are claimed to do, at which point the process starts from the beginning.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 15:18 UTC (Wed) by halla (subscriber, #14185) [Link] (7 responses)

Yeah, right. Hyperbole is a form of art, but much overrated. But don't let reality stand in the way of your wit: why should it? I bet your sharp analysis made you feel good.

But I've been a long term KDE user and a developer since 2003. And while ther definitely is developer turnover, as in any healthy project, I see a lot of familiar people committing daily -- some of them have been at it for over ten years. Technologies like KIO have been around for that long; other technologies, like dcop, have gotten phased out. But not because people lost interest: because Gnome refused to use it as well, and it was replaced by a freedesktop standard, dbus. And some things, like the kicker, have been replaced wholesale. But not because the maintainer had disappeared, but because he had learned and felt he could do better. And he did -- much better, by any metric that counts: crashes, new contributors, flexibility, prettiness. Probably even cyclomatic complexity...

And, well, I'd like Nye to prove that he was laughed out of the room when he pointed out small niggling details. I know I fixed today a small niggling detail for a user who approached me about it on irc, and I know that that is far more common than laughing people out of the room.

But nobody likes being condescended to, and if you act like that, don't be surprised if volunteers volunteer their labor towards fixing someone elses bug.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 16:12 UTC (Wed) by anselm (subscriber, #2796) [Link] (5 responses)

It seems to me that the single most important thing KDE could do to counter the impression of leaving crucial bits of functionality hanging in the air is to come up with a credible replacement for »kprinter«. In KDE 3.5, kprinter was an incredibly convenient way to print stuff, and for some inscrutable reason it has been mostly replaced by the vanilla Qt print dialog in KDE 4, which as a piece of software is, by comparison, probably someone's idea of a practical joke (and not a very funny one at that).

A colleague, who is a KDE project member and usually well-informed about what goes on inside KDE, told me that nobody dared to take on porting kprinter to KDE 4 »because the code is so crufty«. Now I've only been in software development for 25 years or so, and it seems to me that a glorified GUI for printer options wouldn't exactly come under the heading of rocket science (not like, say, OS kernels or C++ compilers or stuff like that), but what do I know. What I do know is that I would really like to update some people I support to KDE 4, but since some things that kprinter does are part of the killer feature set they're using daily on their Linux boxes, it's KDE 3.5 for them until someone comes up with a kprinter workalike for KDE 4 – simply because I would find it very difficult to explain to them why the program they have been relying on for so long has gone, with no replacement forthcoming.

And don't get me started on SSL support in Konqueror. The last time the issue came up people told me not to use Konqueror in the first place, which I think is silly; it's a reasonable browser if one doesn't need SSL, and it is well-integrated with the rest of KDE, which is nice. If the KDE project doesn't want people to use Konqueror they should at least be honest and get rid of it altogether, the way they did with kprinter.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 21:12 UTC (Thu) by nix (subscriber, #2304) [Link] (2 responses)

And don't get me started on SSL support in Konqueror.
What's wrong with it? It works perfectly well for me (though admittedly the most I use it for is bugzilla reports and a little online banking).

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 21:56 UTC (Thu) by anselm (subscriber, #2796) [Link] (1 responses)

SSL support in Konqueror in KDE 4.x is mostly OK if all you want to do is look at HTTPS web sites and (like most people) aren't too bothered about certificate authenticity, but during the 4.x transition the KDE folks in their infinite wisdom saw fit to remove everything to do with certificate management. This means you don't get to use things like client certificates, which unfortunately renders Konqueror unable to, say, access our company's internal website. And since the KDE people don't seem to believe in Konqueror as a web browser any longer, there do not appear to be any plans to bring proper SSL support back. (They're probably waiting for someone to implement a new web browser from scratch.)

In practice it is not that big a deal since the SSL support in Firefox is vastly superior to that of KDE 4.x's Konqueror. However, in KDE 3.5, Konqueror's SSL configuration was in turn a lot better than that in Firefox today, and it is simply gone. As I mentioned before, if the KDE project wants us to use Firefox instead of Konqueror they should say so – but Firefox, while a good web browser, isn't nearly as well-integrated with the rest of KDE as Konqueror is. I remember when Konqueror was touted as one of the big pillars of KDE, but now the project seems to be busy rewriting history and downplaying the need for a capable KDE-based web browser in the first place.

Just for the record, I don't have an issue with KDE gaining exciting new capabilities. I've been using KDE almost since it came out and will probably go on using it for all its faults – if only because the competition is even worse. But what bothers me is existing capabilities going away for no good reason whatsoever and without warning.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 22:36 UTC (Thu) by nix (subscriber, #2304) [Link]

Ew. I hadn't noticed that was gone. Yes, you're right, that is bad. :/

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 31, 2010 14:03 UTC (Sat) by kreutzm (guest, #4700) [Link]

Second. This is why my employer now heavily sponsores activities for a new printer dialogue for KDE 4. When we switched from Windows, kprinter was basically a step back already (which caused haeavy uproar), dropping even kprinter would definitly hurt even more. Our aim is now to have a KDE 4 printing solution *on par* with our previous printing solution. And it is based on previous work and will be available under typically KDE conditions.

Dirk Hohndel at Akademy (KDE.News)

Posted Sep 2, 2010 18:15 UTC (Thu) by Baylink (guest, #755) [Link]

Or, perhaps, they could just stop leaving critical bits of functionality hanging in the air?

I've always vastly preferred KDE to Gnome, precisely because KDE seemed as if it had a chief architect who actually understood the target audience's needs; I could install it on machines for business clients, and not have to make excuses.

It would appear that approach has gone entirely in the toilet for KDE4; aside from forcing a complete change of desktop paradigm with Plasma for which my clients are completely uninterested in investing the payroll funding to train, it's not possible to go *back* to an interface design their employees *do* understand -- and worse, as of SuSE 11.3, you can't even install KDE3.5 (at all, so far as I can tell; there are instructions on how you can supposedly do it, but they don't work for me, and I'm not exactly a newbie).

So I guess perhaps Linux really has reached it's zenith, as some people have been suggesting, and is now on the long slow slide to mediocrity -- which is to say, irrelevance, given the monopoly Microsoft already holds on the desktop.

I don't know what other conclusion to draw, when two Open Source people with jobs as high-profile as the Release Configuration Manager for OpenSUSE, and the chief architect for KDE4, drop the ball this badly.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 16:22 UTC (Wed) by nye (subscriber, #51576) [Link]

>And, well, I'd like Nye to prove that he was laughed out of the room when he pointed out small niggling details. I know I fixed today a small niggling detail for a user who approached me about it on irc, and I know that that is far more common than laughing people out of the room.

I'd like to retain some semblance of pseudonimity here, so I'm going to decline. It's understandable if you choose to discount it in those circumstances. I will also add, having gone back to look at some past conversations, that the disdain that really put me off doesn't appear to have come from KDE developers (at least not any of note), but rather from overzealous users.

>But nobody likes being condescended to, and if you act like that, don't be surprised if volunteers volunteer their labor towards fixing someone elses bug.

Bear in mind that this does go both ways - it can take a lot of time to produce a detailed, accurate, and reproducible bug report, and seeing it shrugged off rudely is not exactly inspiring.

It's unfortunate that the goals of the KDE project have also shifted somewhat - it can be disheartening that a feature regression in some component that was rewritten from scratch is first labelled as WISHLIST (because the component is new, so missing functionality is not technically a bug) and eventually closed WONTFIX (because nobody capable of implementing it wants that particular feature). This is largely unavoidable no matter how politely the refusal is couched. Sadly several things which to me made KDE stand out from the competition fall into this category.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 13:39 UTC (Thu) by jospoortvliet (guest, #33164) [Link]

Wait. You're surprised to find keybindings under the "Keyboard and Mouse" section in the configuration screen? Where else would you expect them to be?

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 20, 2010 21:43 UTC (Tue) by cmccabe (guest, #60281) [Link] (17 responses)

If you're looking for a simpler alternative to Gnome, try Xfce. I've been using it for a few months and it's awesome.

(I have friends who use KDE but it's a little heavyweight for me.)

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 21, 2010 20:17 UTC (Wed) by oak (guest, #2786) [Link] (16 responses)

I recently installed Xubuntu[1] for a relative and Xfce is indeed really nice.

About only complaint I have about it is its filemanager having no convenient way to access remote files and some gaps in translations. Fuse support will apparently come only in the next version.

[1] Kubuntu KDE was way too bloated for 1/2GB of RAM. Ubuntu Gnome has really bad looking theme (at least on cheapo Nvdia + CRT), gnome panel applets are bloated and Evolution is my least favorite mail client (mutt, Thunderbird and KMail are all OKish, Claws/Sylpheed I haven't used enough to comment on).

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 13:37 UTC (Thu) by jospoortvliet (guest, #33164) [Link] (15 responses)

Note you can mix and match things if you need... If the Plasma Desktop is too heavy, feel free to use another one. Doesn't mean you can't use KMail or Gwenview or Dolphin or...

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 17:04 UTC (Thu) by oak (guest, #2786) [Link] (14 responses)

> If the Plasma Desktop is too heavy, feel free to use another one.

So the main memory bloat in (Ubuntu) KDE4 comes from that? What alternatives work OK for KDE?

Unfortunately one cannot install them on (installation) live-CD for testing. It was easier just to pop in another live-CD to check out next desktop alternative...

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 17:17 UTC (Thu) by jospoortvliet (guest, #33164) [Link] (13 responses)

You tell me where it comes from. I've heard a lot of ppl complain, but haven't seen any numbers, and for me it ain't that bad.

Surely for sub-512 mb a modern KDE set of apps with plasma and akonadi (in it's current form) and Nepomuk can be heavy. There is a lot of optimization work still to do (and going on). You simply can't expect to have 2010 functionality with win 95 system requirements...

But a lot of work is going on to decrease resource usage, and for a system which provides as much as a full blown plasma desktop + kde apps it doesn't use much resources, I would say...

Many of the new infrastructural bits like Akonadi are meant to lower resource usage - eg Akonadi will make sure only one copy of your addressbook has to be in memory instead of several like in the pre-akonadi KDE PIM apps. Nepomuk shares searching and indexing between apps (does that job for Akonadi, for example) and hopefully at some point we will have a common DB solution as well.

Yes, there is work to do, no, compared to technology with equivalent functionality KDE apps are NOT big or resource-hungry. Despite the fact there is room for improvement, and despite the fact we suffer from stupidity in the lower stack (eg ANY app using OpenGL with NVidia incurs a 10 MB runtime memory overhead due to stupidity in the proprietary driver).

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 22, 2010 18:04 UTC (Thu) by oak (guest, #2786) [Link] (12 responses)

> You tell me where it comes from. I've heard a lot of ppl complain,

The main issue was slowness and as KDE used more memory than others, I assume slowness to come from that.

This was on live-CD, so the effect of extra memory usage was worse than it would be with an installed distro. For example the libraries used by KDE are larger than Gnome ones (e.g. Qt v4 is huge compared to Gtk v2) so there will be more code page paging from CD.

> but haven't seen any numbers, and for me it ain't that bad.

PSS from the /proc/PID/smaps data is nice for investigating relative heaviness between apps. Here's an UI for it:
http://www.selenic.com/smem/

xrestop can be used to check which clients use most X resources:
http://freedesktop.org/wiki/Software/xrestop

For tracking resource leakage (not just memory), you might try this Maemo tool:
http://wiki.maemo.org/Documentation/devtools/maemo5/sp-en...
http://wiki.maemo.org/Documentation/devtools/maemo5/sp-en...
http://repository.maemo.org/pool/fremantle/free/s/sp-endu...

> You simply can't expect to have 2010 functionality with win 95 system requirements...

Win95 worked with 64MB of memory, this machine had 512MB. But I wasn't expecting anything particular, just checking which Desktop was most suitable for that machine and my relative. It was Xfce.

Surely you're not expecting somebody to buy a new expensive computer to compensate for KDE slowness when he just wants to upgrade an older nicely working distro (with KDE3) before its security updates stop..?

> For a system which provides as much as a full blown plasma desktop + kde apps it doesn't use much resources

Well, for me and my relative eye-candy just isn't worth the cost and I've never seen actual benefits from anything marketed as "semantic". Maybe that's useful for some people, but I don't seem to belong to that group (I have music collection still on CDs etc).

> Akonadi will make sure only one copy of your addressbook has to be
> in memory instead of several like in the pre-akonadi KDE PIM apps.

Does that one copy use less memory than the multiple instances with previous KDE version? :-)

Btw. KDE switched from DCOP to D-BUS. Are you happy with that choice (performance etc)?

> eg ANY app using OpenGL with NVidia incurs a 10 MB runtime memory overhead due to stupidity in the proprietary driver

NVidia driver isn't the only driver having this kind of issues. If the 3D graphics chip & driver don't handle textures, GL command buffer etc being swapped out, they need to be locked in RAM which makes memory much more tight for everything else...

Compositing, double/triple buffer etc all consume a lot of memory. If one cares about eye-candy, it can be a good trade-off, but if one doesn't, it's a pretty huge cost for a fairly small usability advantage (shadows, live window thumbnails/switching, tearing elimination, panning speed improvements useful on touchscreens etc).

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 23, 2010 19:40 UTC (Fri) by jospoortvliet (guest, #33164) [Link] (11 responses)

This was on live-CD, so the effect of extra memory usage was worse than it would be with an installed distro. For example the libraries used by KDE are larger than Gnome ones (e.g. Qt v4 is huge compared to Gtk v2) so there will be more code page paging from CD.

True, KDE apps share more code through the libs. That also means they use less memory if you run them together. So for running one app, KDE apps are inefficient. Having a full set of apps running - like 5-10, and the sharing of functionality becomes really nice.

xrestop can be used to check which clients use most X resources:

That one is fairly interesting, showing here Plasma using 21 mb. It also uses about 25 mb on it's own, making 50 mb in total... Biggest mem eaters are (as usual) webbrowser and mail client. Both go over 100 mb here...

Surely you're not expecting somebody to buy a new expensive computer to compensate for KDE slowness when he just wants to upgrade an older nicely working distro (with KDE3) before its security updates stop..?

That sucks. Then again, anything sold in the last 5 years can run a modern KDE plasma desktop pretty well... I have a 1.2 ghz dualcore 1 gb ram laptop - doing great. I don't see why it's an issue to run eg IceWM on hardware not capable of running the latest KDE software - but we aim for slightly more up to date hardware. We wouldn't want the whole FOSS stack to be held back just to cater for old hardware?!?

Well, for me and my relative eye-candy just isn't worth the cost and I've never seen actual benefits from anything marketed as "semantic". Maybe that's useful for some people, but I don't seem to belong to that group (I have music collection still on CDs etc).

Again a valid reason not to use it. Part of the problem here is of course that the tech simply isn't completely mature, and more importantly, integrated in apps yet. Bangarang is a positive exception here - it is very small in itself, using high-level KDE tech to do its thing (nepomuk to find, index and display music, Phonon to play it etc). This keeps the work for the dev minimal, apps small and maintainable etc.

Does that one copy use less memory than the multiple instances with previous KDE version? :-)

It should :D

Btw. KDE switched from DCOP to D-BUS. Are you happy with that choice (performance etc)?

Hmmm, it's still not as fast as DCOP but getting there afaik.

Compositing, double/triple buffer etc all consume a lot of memory. If one cares about eye-candy, it can be a good trade-off, but if one doesn't, it's a pretty huge cost for a fairly small usability advantage (shadows, live window thumbnails/switching, tearing elimination, panning speed improvements useful on touchscreens etc).

Well, in some cases these improvements are very important - the Plasma Netbook interface for example makes heavy use of compositing, and so does more and more software. I think, again, that we shouldn't be held back by the existence of old hardware. Besides, my GMA950 runs compositing just fine - and that video chip is truly amazingly sucky...

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 24, 2010 10:30 UTC (Sat) by oak (guest, #2786) [Link] (6 responses)

> Having a full set of apps running - like 5-10, and the sharing of functionality becomes really nice.

Who runs this many separate apps at the same time? Many windows yes, but apps? Or do you count also always running system services as apps?

>> xrestop can be used to check which clients use most X resources:
>
> That one is fairly interesting, showing here Plasma using 21 mb. It also uses about 25 mb on it's own, making 50 mb in total...

AFAIK xrestop doesn't even show all of the client's X memory usage. I think for example some of the memory used for compositing can be missing.

> Biggest mem eaters are (as usual) webbrowser and mail client. Both go over 100 mb here...

Browser has good reasons for huge memory usage, but I don't see why mail client should eat so much memory, it doesn't need to keep the mail contents in memory, and not even all of the headings. That's what databases are for...

> Then again, anything sold in the last 5 years can run a modern KDE plasma desktop pretty well...

That machine was about 5-6 years old (mine is 7 and works fine), the problem is that he had bought it with 256MB RAM (with Windows) which I have upgraded to 512MB RAM (with Linux dualboot) which is fine as long as one doesn't use GL-desktops.

> We wouldn't want the whole FOSS stack to be held back just to cater for old hardware?!?

We're not asking for that, we're just happy using Xfce. :-)

I think the issue is mainly about RAM, not the age of the computer[1]. One needs minimum of 1GB RAM for KDE4. For XFce and current Gnome (without compositing) 1/2GB is fine.

It's required mainly for Browser (and e.g. photo editing), modern www-pages don't work really well with 256MB of RAM when one's also running a real desktop. With IceWM and stripped down system services 256MB would probably still be OK for browsing, at least if one has fast swap...

[1] except that all new computers are now for Windows 7 which also needs at least 1GB.

>> Does that one copy use less memory than the multiple instances with previous KDE version? :-)

> It should :D

Next question is then, were those three separate instances usually in the RAM at the same time in earlier KDE version (i.e. did most users run corresponding apps at the same time)? If not, it means that the KDE base memory usage has grown also in this respect for that set of users.

(Common functionality has of course also other benefits, I'm not saying something is better or worse, just pointing out memory usage implications.)

> Besides, my GMA950 runs compositing just fine - and that video chip is truly amazingly sucky...

Compositing is mostly about pretty simplistic operations on very large amount of texture data. I.e. memory and memory bandwidth is an issue, not how much 3D power the chip has.

GMA950 is integrated chip using system memory right? When you're compositing non-GL applications, that's not a bad thing because they're updating that memory on the CPU side. Only for GL using apps having these buffers on chip's faster memory and compositing them there would be faster.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 24, 2010 12:51 UTC (Sat) by jospoortvliet (guest, #33164) [Link] (5 responses)

> Who runs this many separate apps at the same time? Many windows yes, but apps? Or do you count also always running system services as apps?

I currently don't have a lot running here - a mail app, a calendar app, a news feed reader, a webbrowser, a chat application. That's five, and about my minimum. I often also have (on other virtual desktops) more browser windows, text editor (kate or kwrite), a konsole or two, and often a video player window, filemanager, word processor, video editor and more random crap.

I personally like and thus use the session management in plasma desktop - I don't close everything all the time if I know I'll work on it later today or tomorrow. So I often have over 10 apps open on different desktops. Linux/KDE handles it just fine. Sure win XP wouldn't but that ain't a multi tasking OS anyway.

In terms of resources, I'm now behind my laptop and it uses 673 mb ram according to free. That's with kmail running with about 100.000 mails in 15 folders and 15 tabs open in Rekonq. And thousands of articles in Aggregator. Yes, that's indeed over 512 mb, I think you're right in that a full-blown plasma desktop + set of apps running needs 1 gb. I often edit video on this thing, and it works great, so that 1 gb is also enough for most purposes. Though I must admit, behind my main workstation with 4 gb ram I often manage to fill up 2 GB by not closing apps and having 6 virtual desktops full of stuff ;-)

And yes, KMail eats a lot of ram, I hope Kmail2 will be better in that regard.

> Next question is then, were those three separate instances usually in the RAM at the same time in earlier KDE version (i.e. did most users run corresponding apps at the same time)? If not, it means that the KDE base memory usage has grown also in this respect for that set of users.

Yeah, they did. If you had contact running with the calendar, addressbook and email app, the address book was actually running three times: by itself, within KMail and within the calendar app. Such things happened rather often, and it's now completely solved.

But even if it doesn't save memory (and it only does that in more involved usage scenario's, not in a more minimal usecase) it helps developers write better apps as I wrote earlier - important if you're not a big corporation but a bunch of volunteers with (sometimes high) turnover...

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 26, 2010 11:11 UTC (Mon) by nye (subscriber, #51576) [Link] (4 responses)

>Sure win XP wouldn't but that ain't a multi tasking OS anyway.

Why do we always have to get this kind of mudslinging here? Windows XP does that just fine, and because it uses a fraction of the RAM running background processes compared to KDE4, it frequently does it better.

The biggest problem with multitasking in XP is that in certain scenarios under load (eg. trying to launch large applications at the same time as running a batch file copy) it has poor IO scheduling behaviour, which is a little better in Linux (and can be made substantially better with a lot of careful tweaking).

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 26, 2010 11:26 UTC (Mon) by jospoortvliet (guest, #33164) [Link] (3 responses)

KDE ain't an OS, it can run on top of Windows XP. But in terms of multitasking there is more wrong in XP than just the bad IO scheduler - the CPU scheduler also sucks, as does networking with too many processes.

And so does the multi-user separation - eg running a single process as root is difficult. Besides, it manages to take longer here to start up on the same hardware as I have at home - which is quite an accomplishment as my home computer also restores the 15+ apps I have running there while rebooting, while XP does not have such basic functionallity.

Then again, maybe it IS mean towards XP to expect anything from it - it's almost 10 years old after all. But don't think my complaints are based on some bad rep the OS has with me - I use it every day at work and I hate it with a deep passion. For the simple reason it sucks in so many ways... It might not crash often, but it is inconsistent, overly complicated, undependable and annoying in so many ways it's unbelievable people put up with it.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 26, 2010 15:36 UTC (Mon) by nye (subscriber, #51576) [Link] (2 responses)

>KDE ain't an OS, it can run on top of Windows XP.

[This paragraph is really an aside]
Indeed. Although it's not really there yet, and consumes far more resources than the default shell (kded4.exe and klauncher.exe between them use more RAM than this explorer instance that's been running for the past fortnight, and they don't even *do* anything so far as the user is concerned). I'd say that the memory consumption of KDE applications, plus the slightly increased CPU use coming from having so many abstraction layers on top of each other, leads to a generally poorer experience when using a lot at once. This is particularly the case on low- to mid-range hardware, and more than offsets any real differences between the multitasking capabilities of the underlying OS.

>But in terms of multitasking there is more wrong in XP than just the bad IO scheduler - the CPU scheduler also sucks,

All CPU schedulers suck, but this is a red herring. There's no noticeable difference in any standard desktop usage pattern between XP and stock Linux CPU schedulers. Both bounce processes across cores constantly, and both are poor for realtime interactive use, but if you're not doing low-latency audio processing or the like then it's basically irrelevant.

> as does networking with too many processes.

This sounds highly plausible.

>And so does the multi-user separation - eg running a single process as root is difficult. Besides, it manages to take longer here to start up on the same hardware as I have at home - which is quite an accomplishment as my home computer also restores the 15+ apps I have running there while rebooting, while XP does not have such basic functionallity.

All entirely fair.

>Then again, maybe it IS mean towards XP to expect anything from it - it's almost 10 years old after all. But don't think my complaints are based on some bad rep the OS has with me - I use it every day at work and I hate it with a deep passion.

As do I, but despite a deep loathing for it, I nevertheless use it on my home desktop because it provides the least execrable experience I've found since Debian dropped KDE3. Having been using it exclusively for the last 6 months, and occasionally before that, I've really come to dislike people slamming it where it is not due *because I want to be taken seriously in the cases where it is, rather than dismissed as yet another zealot*.

Simply stating that XP is not a multitasking OS is absurd, as are numerous other dubious claims levelled against it, and throwing slurs around haphazardly rather weakens the argument in the cases where it behaves in such a brain damaged manner that the developers responsible deserve to be rounded up and shot.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 26, 2010 15:48 UTC (Mon) by jospoortvliet (guest, #33164) [Link] (1 responses)

I think we're very much in agreement, even though I can and do use a Plasma desktop at home (4.4.4 right now) and it's fine. If only the bloody Intel drivers for Xorg didn't start to crash - since 2 weeks they often do.

About the KDE-on-Windows memory consumption, I'm unsure if it's due to the abstraction layers (those pretty much go away once you compile and run the app, afaik - but I dunno if Windows has an equivalent of mmap, which might explain the problem). It's probably also more the very new nature of our support for this platform, the fact we duplicate lots of things instead of using native functionality, the mediocre compiler etc. IOW I doubt it can't be fixed by trowing some resources at it - we just don't have those. Apparently there are a lot of KDE-on-Windows users but it hasn't lead to any new developers yet.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 26, 2010 16:43 UTC (Mon) by nye (subscriber, #51576) [Link]

>I think we're very much in agreement,

Okay, sounds good.

>Apparently there are a lot of KDE-on-Windows users but it hasn't lead to any new developers yet.

I suspect there are very few people with both the relevant skillset and motivation. I for one find development on Windows to be an excruciating process, but those with substantial Windows development experience mostly seem to find it the other way around.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 24, 2010 18:30 UTC (Sat) by nix (subscriber, #2304) [Link] (3 responses)

Well, in some cases these improvements are very important - the Plasma Netbook interface for example makes heavy use of compositing, and so does more and more software. I think, again, that we shouldn't be held back by the existence of old hardware. Besides, my GMA950 runs compositing just fine - and that video chip is truly amazingly sucky...
So what do we do if we want compositing and a flexible window manager? kwin has awesome effects but is simply too inflexible for me to tolerate :/ there appear to be no window managers that can both composite and do anything on the customizability front.

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 25, 2010 9:36 UTC (Sun) by jospoortvliet (guest, #33164) [Link] (2 responses)

Dude, what do you miss in KWin in terms of flexibility?

It has tiling
It has tabbing
It has fully configurable window decoration buttons
It has loads of configurable compositing plugins and you can write more
It has very flexible special window settings (giving each window custom properties, finding them based on regular expressions)
It can be scripted (http://rohanprabhu.com/?p=74)

What else do you need?

Dirk Hohndel at Akademy (KDE.News)

Posted Jul 31, 2010 20:30 UTC (Sat) by nix (subscriber, #2304) [Link] (1 responses)

The scripting is new and I had no idea about it :)

I moaned about this a few months ago: http://lwn.net/Articles/383746/. Basically I'd like an equivalent of fvwm's 'maximise on keystroke', 'maximise until you hit things on the same layer as you' and something like layering (so I can push some floating windows above all other windows and back down again with a keystroke).

If nobody else fixes this fixing it is on my todo list because I just can't stay with fvwm forever.

Dirk Hohndel at Akademy (KDE.News)

Posted Sep 4, 2010 19:11 UTC (Sat) by nix (subscriber, #2304) [Link]

In fact, the scripting barely *exists*: it's still being implemented. There is zero documentation besides blog posts (at least, none that google finds). I can't tell what version it is implemented in: it looks like trunk only.

Suggesting I use this barely-existing ultra-bleeding-edge feature to fix a problem *now* is more than slightly unhelpful :/ perhaps in KDE 4.6, maybe, but until then it looks like I'm stuck with fvwm. (Again, 'raise these windows to the top when a key is hit and stop them being obscured, lower them again when another key is hit' is not rocket science: but it seems to be something KWin is currently incapable of.)


Copyright © 2010, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds