Weekly Edition
Return to the Press page
| |||||||||||||
How Desktop Linux Should Behave (OSnews)
OSnews is carrying a
list of complaints about the X desktop which appears to have been
inspired by last year's distributions. "Please support the Y Window
System. There's no fixing X11 that doesn't involve superhuman genius
hackery. Workaround after workaround will only make it more big and
bloated. X11 must be retired. Y Windows is a natural choice since the
project plans X11 compatibility to ease migration." (Thanks to Jay
R. Ashworth).
(Log in to post comments)
Bloated X? Posted Jul 28, 2004 14:24 UTC (Wed) by rsidd (subscriber, #2582) [Link] Where does this "bloated" idea come from? If X wasn't too bloated for a 1980s machine with a 20 MHz processor and 16 MB of RAM, it's not too bloated for today's machines. As for the other stuff -- alpha blending, etc -- it's all being done by the xorg folks, and I dare say their CVS tree is already more usable than "Y-Windows". (And this article is annoyingly short on what else is lacking in X today -- my suspicion is, "very little".)But then, how seriously can you take someone who complains about how hard it is to install anti-virus software on Linux?
Bloated X? Posted Jul 28, 2004 15:37 UTC (Wed) by oak (subscriber, #2786) [Link] Heh, I laughed aloud when I read the Y window system paper where the author said that with *a couple of weeks* of effort it could be ready for real use.Without lots of real applications (browser, office suite etc) and good performance (*lots* of hardware drivers) it will not have any real use. I doubt both of these goals could be achieved by emulating X11 libs, at least in a way that user would get some real benefit out of it. Technically the paper seemed excellent, it was just the effort estimates, schedules and competiveness of a new (any) window system which were funny (I've followed MGR, GGI, Berlin/Fresco, Microwindows, PicoGui etc development). X11 Composite extension basically does what the author describes as the main thing i.e. it doublebuffers top level windows. W window system did that already 10 years ago, although it was otherwise technically inferior. - One of the W Window System[1] authors [1] Not the X predecessor one...
Anti-X Idiots Posted Jul 28, 2004 14:51 UTC (Wed) by elanthis (guest, #6227) [Link] Here's a hint: the author doesn't understand anything about X11 or Y-Windows. How can you tell? He suggests that X11 must be replaced by Y-Windows in order for the desktop to be worth anything.X11 is perfectly fine. The actual protocol is damn near perfect. There is no real bottlenock. It's as fast as you can possibly get when running on your local machine; it'll use the fastest IPC mechanism it can completely transparently to the enduser or application. The features of X11 can easily be extended using extensions, which is exactly how X11 has grown over the years. Font system sucked, so we added Render and client-side font libraries, and the problem is basically solved. Program startup is slow due to initial setup round-trips, so a new extension is being designed to do all this in a few commands vs the hundreds currently used; old apps will still work with no modifications (like switching to Y-Windows would need, without an X11 proxy on top of it) and new apps will just load much faster. Want transparency? Just use Composite. etc. These extensions do not in any way indicate an actual problem with X11. Extensions don't require any special considerations or massive resource loss or anything like that. Adding an extension is basically the exact same as upgrading the core protocol, except that with an extension the feature is optional; that means clients can detect if the extension is there, and if not, simply not use it, making the app backwards compatible with servers many years old. Switching to Y-Windows wouldn't give us any features X doesn't have or isn't capable of having (and will have very soon, thanks to the work at Xorg and Kdrive), while it _does_ break any semblence of backwards compatibility, and itself will just need modifications or extensions in the future for all the stuff we might want that the Y-Windows developers didn't think of today. With even a basic understanding of X11, the above is fairly clear. It's hard to take OSNews seriously when they keep letting uninformed morons post articles like this that serve no purpose at all other than to make it clear how ignorant the author is.
Anti-X Idiots Posted Jul 31, 2004 4:17 UTC (Sat) by dbrandon (guest, #1090) [Link] Well, I know nothing about X-Windows internals myself, but the complaints that people who dotend to be about
- labyrinthine code base (not surprising given X's age)
Sure, the OSNews guy is an idiot. The problems programmers tend to talk about aren't about the
However, it seems from the OLS report that they're working on fixing at least the code base.
But whatever.
Anti-X Idiots Posted Aug 2, 2004 0:13 UTC (Mon) by daenzer (✭ supporter ✭, #7050) [Link] > - labyrinthine code base (not surprising given X's age)
The ongoing modularization efforts should help with that.
> - freakish API and event system (papered over by Gnome and KDE, but not fixed)
I assume that this is about Xlib (libX11), which isn't X itseld but only basically a wrapper library for the protocol. Its major issues should get fixed by another such library being developed called XCB (on top of which an Xlib compatibility library can be used).
> - sheer size
The size of the monolithic tree is mostly due to the large number of components it contains, so modularization should help with that as well.
Can we ignore OSNews Posted Jul 28, 2004 15:41 UTC (Wed) by paulj (subscriber, #341) [Link] I dont read LWN for links to drivel, and I have never read anything on OSNews that did not seem as if it were written, as a previous poster but it, by uninformed morons. Please stop posting OSNews links :)
Can we ignore OSNews Posted Jul 28, 2004 15:59 UTC (Wed) by huffd (guest, #10382) [Link] I concur, even articles LinuxToday posts from OSNews seem to be inflamitory, LWN is doing it for the same reason, to drive up the hit count.Here in the states we call it Yellow Journalism, which stemmed from a 1895 New York comic strip. Next time you see OSNews in the credits just tell yourself it means "go past" or "get beyond", we've all given them enough mind share.
Can we ignore OSNews Posted Jul 28, 2004 16:05 UTC (Wed) by corbet (editor, #1) [Link] "LWN is doing it for the same reason, to drive up the hit count."You do realize that the only thing a bigger hit count means for us is increased bandwidth costs, right? If we wanted to inflate hit counts, there's no end of stuff we could post that would generate thousands of comments and generally keep the site busy. But that is of no real interest. I'll confess to posting things that, one hopes, might increase subscription counts, but not hit counts. I posted that item as a data point on how the Linux desktop is perceived in some quarters. Besides, news is a little thin this morning. In retrospect, perhaps it wasn't worth the effort, but so it goes.
Can we ignore OSNews Posted Jul 28, 2004 16:46 UTC (Wed) by dmarti (subscriber, #11625) [Link] You have category filtering -- why not just make a "Trolls, FUD and Drivel" category and let subscribers choose?
Can we ignore OSNews Posted Jul 29, 2004 7:27 UTC (Thu) by sidboyce (guest, #10891) [Link] Seems a sound suggestion to me. It's obvious the author of the piece was either trolling or knows zip about Linux. Trolling is most likely as the article described Linux going back to about 1995, though it's possible he wrote the article back then, but it has now only been accepted for publication or he was comatose until now. OSNews does not cater for responses to their articles, probably for obvious reasons.
Can we ignore OSNews Posted Jul 28, 2004 18:07 UTC (Wed) by mmarsh (subscriber, #17029) [Link] I'd just like to say that I feel it's entirely appropriate to post links to articles like this occasionally. While I'm certainly not interested in seeing All FUD, All The Time (not that this really counts as FUD, per se), it's good to see as many positions as possible. If we start ignoring obvious trolls and FUD, we start ignoring SCO press releases, Halloween memos, and so on.Posting and commenting on articles like this has the added advantage that the comments teach others both some of the finer details relating to an issue and how to respond to certain criticisms or outright attacks.
Can we ignore OSNews Posted Jul 28, 2004 21:33 UTC (Wed) by AnswerGuy (guest, #1256) [Link] Jonathan,Please don't be too concerned about the votes to exclude OSNews from linking. If it's relevant to Linux and will be read by a significant number of other people out there (including *prospective* Linux users) then link to it. Expressing some healthy criticism or skepticism in the blurb that points to it should be enough to steer the impatient away from ignorant ramblings. Besides, the link might inspire some to recommend sites that refute or propose solutions or alternatives to FUD claims. JimD
Can we ignore OSNews Posted Jul 28, 2004 16:23 UTC (Wed) by paulj (subscriber, #341) [Link] I disagree strongly with your opinion on LWNs' motivations for posting OSNews links, but yes, I should just ignore OSNews articles myself - just so bloody difficult by virtue of how crap they are! ;)
Can we ignore OSNews Posted Jul 28, 2004 16:05 UTC (Wed) by amk (subscriber, #19) [Link] Agreed; OSOpinion articles are similarly clueless. They mean well, but it's apparent the authors aren't programmers and have no idea how complicated it can be to tie things together. (e.g. from the linked-to article: "In other words, when a KDE app is run in GNOME, it should act and behave like a GNOME app and vice versa. ... I'm sure this can be done.")
Can we ignore OSNews Posted Jul 28, 2004 16:19 UTC (Wed) by khim (subscriber, #9252) [Link] It just can not understood what he want! Think about this sample. "When a KDE app is run in GNOME, it should act and behave like a GNOME app and vice versa". Can it be done ? Absolutely. How it can be done ? With centralized development of both. That's how "Cocoa and Carbon applications interact". But why the heck we'll need them both then ?
How Desktop Linux Should Behave (OSnews) Posted Jul 28, 2004 17:14 UTC (Wed) by ccchips (guest, #3222) [Link] I'm tired of these articles, but then I guess they do strengthen people's resolve to defend X11 against nonsense.I think any GUI that inherits Linux from X would probably be some kind of fork from X. I think it's too good to give up, and nobody has shown me otherwise yet.
How Desktop Linux Should Behave (OSnews) Posted Jul 28, 2004 20:03 UTC (Wed) by cpm (guest, #3554) [Link] I completely agree. I didn't start using X daily until 1995.When I finally *got it*, I just couldn't see using anything else. It did take me a while to understand what was going on, but
How Desktop Linux Should Behave (OSnews) Posted Jul 28, 2004 21:49 UTC (Wed) by piman (subscriber, #8957) [Link] Incidentally, there already is a Y Window System. http://www.hungry.com/old-hungry/products/Ywindows/Y.html is 6 years older than this project.
More Ideas #1: LVM! Posted Jul 29, 2004 0:10 UTC (Thu) by AnswerGuy (guest, #1256) [Link] As another comment on the original article, when he talks about "More ideas" and asks for a way to dynamically resize filesystems by adding additional drives ... he clearly doesn't understand that we already have it, and we've had it for a few years.Yes, it's called LVM. (We also have LVM2 and EVMS under development). So I can have /home, /var, and /opt (for example) as logical volumes (presumably in a volume creatively named something like "vg01" or "Volume00"). I can mark partitions, whole disks or groups of disks (metadisk RAID sets using the Linux md drivers) as physical volumes (PVs) and add them to my volume group(s). (I can have multiple VGs, but I don't know why I'd want to). Then I can create logical volumes (LVs), sizing them to be fair small at first and leaving some space available/unallocated in the VG. Using lvcreate makes a volume group and puts a device node under /dev/vg01/ (or /dev/Volume00/ or whatever). Then I can mkfs (any type of filesystem I like) over that virtual block device. Later if I run out of space on an ext2/3 filesystem I can just use e2fsadm to adjust the LV size and resize the fs. When I see that I'm running out of PV space I can just add another drive and use vgextend to keep more space available. Of course I haven't seen a pretty GUI or webmin module for LVM (though EVMS seems to have GUI and curses front ends). But the point is that we've had logical volume management for awhile, probably longer than that author as been using Linux. It's well supported (and the suite of tools is included on LNX-BBC and Knoppix, it's been offered with S.u.S.E. for years, and support has been hidden under the hood in Red Hat, and Debian, too --- though their installers don't offer it as a suggestion). It doesn't seem to have any noticeable performance impact (though I've only used it on fileserver, desktop and laptop applications --- it's not what I'd use for a web, mail, or database server or in a cluster where I wanted to eke as much performance as possible from the systems). The other things he was asking for probably also exist somewhere to some degree. JimD
Summary: Review of very old Slackware in 2004 Posted Jul 29, 2004 8:43 UTC (Thu) by jmayer (subscriber, #595) [Link] This looks like the review of a very old Slackware (RedHad, Suse,whatever). With all that OS testing, the author probably mixed up his old Linux CDs with the new ones and installed the ones from 4 years ago. I consider it a waste of time to actually refute the points - readers of lwn should know that already :)
Summary: Review of very old Slackware in 2004 Posted Jul 30, 2004 16:05 UTC (Fri) by danielos (subscriber, #6053) [Link] well, I'm using Debian in a laptop starting from woody, and now upgraded to sid (yes, never heard nothing about missing dependencies), and I think author is not so wrong when he talk about hardware problem, I know there are a number of daemon that just do what he say, but you have to know those exist and how are named.Maybe is better if someone indipendent (as gnu) say which daemon/library is the best, how they interoperate, if there is a standard API, and so on. Till now every distro develop his own method for doing hardware discover so there are a number of methods and no unique idea about what exactly means this task. As I think is stupid to trash X window, I think is stupid to trash pcmcia-cs, discovery, bluez and other, but those are not consistent. I'm not talking about user perception: from the design prospective hardware recognition in Linux is crufty. Hardware is a relevant part of an OS, it should have a place in a glibc's addon, like nptl and crypto.
|
|||||||||||||