Weekly Edition
Return to the Front pageSponsored link
Serve your customers, not your servers, with VERIO Linux VPS. Full-access test-drive here.
| Weekly edition | Kernel | Security | Distributions | Search |
| Archives | Calendar | Subscribe | Write for LWN | LWN.net FAQ |
"Just works with Linux"
It's a nice idea, but not a particularly new one. Your editor has seen a fair number of these databases come and go over the last ten years. Starting a "just works" database is easy, but keeping it current and relevant is hard, for a number of reasons:
- The variety of hardware out there is huge. Simply testing and
creating entries for a meaningful subset of the available gadgets is a
major task.
- Vendors feel free to change the internal makeup of their gadgets
without telling anybody - or changing the model number. The changes
in the LinkSys WRT54G router are a recent example. This behavior
complicates the database (which must now have information on telling
working hardware from paperweights) - and its maintenance.
- Nobody can actually have all that hardware around, so information must
come from a wide community. Most of us only buy hardware
sporadically, so we tend to have little motivation to help with the
ongoing maintenance of a hardware database. Some of the information
which is contributed may also be of dubious reliability.
- Companies which might help with the maintenance of such a database
have their own incentives to deal with. Red Hat maintains a hardware
list, for example, but it (1) is small, and (2) talks
about RHEL, not about Linux in general. The company once known as
Linuxcare had the proper motivation to maintain a good list, but,
well, Linuxcare didn't weather the dotcom bust very well.
- Weird factors come into play. The BlueZ project used to have a very nice list of working hardware, but that list was pulled down as a result of objections from the "Bluetooth Qualification Administrator."
Any future attempt to build a Linux hardware compatibility database will have to find a way to overcome the problems listed above. The task is not impossible, but it may well beyond what a volunteer project can sustain. It looks, instead, like the kind of work which can be helped by the addition of a stream of money. Perhaps an industry group (OSDL, say) would like to serve the community by taking this task on.
Meanwhile, your editor notes with dismay an increase in the number of Linux-installed hardware vendors who are shipping systems with proprietary drivers. Once upon a time, the purchase of a system with Linux pre-installed was worth the extra cost just because the running Linux instance was a positive proof that the hardware was, indeed, supported. When these vendors ship non-free "Linux" systems, they violate that guarantee - and destroy much of the value of their product. Unfortunately, "buyer beware" remains necessary advice for those buying hardware to work with Linux.
(Log in to post comments)
MUST note if drivers are OSS - and the FSF has such a list!
Posted Dec 15, 2005 3:39 UTC (Thu) by dwheeler (subscriber, #1216) [Link]
The last portion of the article illustrates a critical part of such a database: does it have open source drivers? If it doesn't, it will become a paperweight sooner or later; no hardware vendor can afford to support old cards indefinitely.One resource you forgot to mention is the Free Software Foundation (FSF)'s GNU/Linux hardware devices list. Like the other lists, it's small compared to the universe of hardware out there (many cards that are supported aren't listed!). And the FSF is a harsh grader: they want 100% Free Software drivers, so partial OSS drivers (common in wireless cards, for example) get no credit. Once you keep that in mind, it's a very helpful resource. I found helpful information about Ralink wireless cards from this site, for example; I'm using one of the cards they recommended right now.
MUST note if drivers are OSS - and the FSF has such a list!
Posted Dec 15, 2005 16:43 UTC (Thu) by vmole (subscriber, #111) [Link]
Perhaps the reason he "forgot" to mention it was that it has only two entries, one of which is over a year old and is marked "unsupported" (I don't know whether that means the SD readers are unsupported under Linux, or that the driver in question is currently unsupported.)
All of which underscores the original point: it's easy to setup the structure for a database like this, and damn hard to make it actually useful.
MUST note if drivers are OSS - and the FSF has such a list!
Posted Dec 15, 2005 19:45 UTC (Thu) by gtaylor6 (subscriber, #19812) [Link]
[...] it's easy to setup the structure for a database like this, and damn hard to make it actually useful.Yup.
I built linuxprinting.org, and setting up the schema, while less trivial than you might think, was by far the easy part.
Gathering the data is hard. You need to be a combination of editor and tech support person to be able to work with inexpert users to determine the status of each new device. We've got roughly 1750 entries, and it took about 10 years. That's less than one device a day, and it's not for lack of printer models to document...
One big problem is that few of these databases share data. If they were all combined, and any submission to any ended up everywhere, then all this discussion about which existing incomplete database should be used would become moot. Unfortunately, seamless data sharing requires a comparatively formal and complex schema that isn't likely to appear out of thin air.
MUST note if drivers are OSS - and the FSF has such a list!
Posted Dec 16, 2005 21:31 UTC (Fri) by liamh (subscriber, #4872) [Link]
I was going to post that linuxprinting is one such database that really works -- there's a large amount of information and it seems to be continually updated. This has been valuable to me several times.
About the FSF effort
Posted Dec 26, 2005 17:39 UTC (Mon) by johnsu01 (guest, #34757) [Link]
I have the responsibility for coordinating the FSF effort at a hardware database. Obviously we have next to nothing there at the moment.
I've been talking with some people over the last couple of weeks about ways to make it easier for people to submit information, and so I think we'll have some improvements in that area soon.
If anyone would like to chip in some volunteer time to help get this thing together, drop me a line at johns at fsf dot org with your ideas. The core principle here is to include only completely free drivers. We will be focusing on GNU/Linux systems, but will be happy to include information about free BSD systems as well when that info comes our way.
can the community support devices "indefinitely"?
Posted Dec 21, 2005 9:12 UTC (Wed) by wilck (subscriber, #29844) [Link]
no hardware vendor can afford to support old cards indefinitely.
Neither can the community. See how many old drivers are being kicked out of the kernel because they're unmaintained or have become non-functional over time? It is a Linux myth that the community supports old hardware forever. All you can reasonably say is that if the driver is free, it remains theoretically possible to port it to recent kernels.
"Just works with Linux"
Posted Dec 15, 2005 6:43 UTC (Thu) by njhurst (guest, #6022) [Link]
There was a list floating around a while back called the linux incompatibility list (http://leenooks.com/) which listed devices that are broken, rather than those that work. I think idea was that as time progress the list got shorter and shorter. Sadly it appears to have prematurely got very short (the name doesn't resolve for me, though google still has a cache).
Linux Incompatibility List - Ooops:-)
Posted Dec 15, 2005 7:34 UTC (Thu) by davidw (subscriber, #947) [Link]
It's still there, but for some reason I didn't get the email saying I needed to renew it, and it expired last night. I have now paid it up for two years and it is back on line!
It is still going strong (unfortunately), and I still think it's a very valid idea, because it best deals with the incentives that Our Editor mentions.
In other words, if stuff "just works", you're unlikely to go poking around finding out what it is and reporting it to a web site. On the other hand, if you discover your brand new Widget 2000 doesn't work with Linux, perhaps you'll be irritated enough to go list it on the Incompatibility List. Also, perhaps people shopping will see which manufacturers are to be avoided, and spend their money somewhere else.
Also, there really is a lot of stuff out there that works with Linux. Much more than hardware that doesn't, so a list of things to avoid is likely to be shorter than a big long list of everything that works fine.
A few more comments
Posted Dec 15, 2005 8:16 UTC (Thu) by davidw (subscriber, #947) [Link]
Turns out the emails warning me were thrown out by Thunderbird's spam filter:-( I'm not really very happy with it (tons of spam gets through, and apparently it loses important emails once in a while), but that's another matter.
One way to make a compatibility list work would be to automate it, to take some of the annoyance out of reporting things.
You'd run a program which would pop up a dialog asking you if it's ok to look at your hardware and send the information back to some centralized database in order to help improve Linux. At that point, it would look through your system, and ask you whether everything seems to be working ok, or where possible, automate testing. That might be sort of tricky, though, especially where less knowledgeable users are concerned...
A few more comments
Posted Dec 22, 2005 17:16 UTC (Thu) by ArsonSmith (guest, #5695) [Link]
I think it needs to be broken down by chipsets that are suppose to be supported in the kernelV2425-3 10-100 ethernet chipset
then by cards that support it
Kellogs ethernet card cheerio net interface netsurfer pro II
then list working status
number of people that have gotten this to work works out of box (no configuration) works after mild configuration requires out of kernel modules to be compiled or patches to kerenl does not work
"Just works with Linux"
Posted Dec 15, 2005 12:25 UTC (Thu) by ekj (subscriber, #1524) [Link]
I disagree. Such a list would be very nearly useless for anyone wanting to buy linux-hardware.The reason is simple: no list is complete, and a list such as this one would likely be less complete than most.
Now, with a positive list: "these devices work:" it is of less consequence to me if the list is complete, if I buy something on the list, and the list-info is accurate (a different problem) then it'll work. OK, so maybe in reality I had a bigger choise than the list gave me, but that's mostly of less consequence, I don't need 10 graphics-adapters anyway, if I get one that works well with free drivers I'm satisfied.
With the negative list, I can't feel secure buying something just because it's not on the list. It could simply be that it's missing from the list because the list is incomplete. (or, like for the other, the list could be inaccurate)
Incompatible List
Posted Dec 15, 2005 15:37 UTC (Thu) by cdmiller (subscriber, #2813) [Link]
There is plenty of usefulness in both the positive and negative lists, I just went and added to the Incompatible list entries for DELL servers and workstations under their Hardware Monitoring entry. Good information to have when looking for Linux servers in particular.
"Just works with Linux"
Posted Dec 26, 2005 22:16 UTC (Mon) by oak (subscriber, #2786) [Link]
> With the negative list, I can't feel secure buying something just> because it's not on the list. It could simply be that it's missing
> from the list because the list is incomplete. (or, like for the other,
> the list could be inaccurate)
It's a fact of life that "Positive list" won't be covering everything and
then when you see a very good deal on something you'd like to have, it's
good to know that it will not work, instead of buying it, testing it and
finding that yes, the reason why it wasn't on the positive list was
because it was broken, not because it was new (product/manufactorer/etc).
"Just works with Linux"
Posted Dec 15, 2005 9:36 UTC (Thu) by MarkVandenBorre (guest, #26071) [Link]
Isn't this what Ubuntu is silently trying to do with its hardware database? Collecting enough info from users for it to be statistically relevant, then through some degree of automation filter out a list of working hardware.
"Just works with Linux"
Posted Dec 15, 2005 17:25 UTC (Thu) by tzafrir (subscriber, #11501) [Link]
Not just Ubuntu. Practically any distro.
"Just works with Linux"
Posted Dec 17, 2005 10:50 UTC (Sat) by csamuel (subscriber, #2624) [Link]
Yeah, but the Gnome version of Ubuntu has the hwdb-client package thatlets you submit your hardware details to a database after running some
tests on each component that you answer to say whether it worked or not.
If (like me) you're using KUbuntu you can do:
$ sudo apt-get install hwdb-client
$ hwdb-gui
To contribute to the database.
Chris
"Just works with Linux"
Posted Dec 17, 2005 11:04 UTC (Sat) by csamuel (subscriber, #2624) [Link]
Although I've just noticed that whilst sound works fine in KDE it doesn'twork in the hwdb-gui app.
Yes, I have disabled arts (the KDE sound server)..
"Just works with Linux"
Posted Dec 15, 2005 10:24 UTC (Thu) by macc (subscriber, #510) [Link]
SuSE had from early on and still has a rather completeannotated and searchable list on
various hardware "Things" ( working and nonworking )
It is rather difficult to find from the Novel site ;-/
Bluetooth vs. BlueZ project
Posted Dec 15, 2005 10:39 UTC (Thu) by walterh (guest, #19113) [Link]
I cannot believe that that the BlueZ project pulled their list because someobscure "license agreement" doesn't allow you to tell which devices work
with Linux.
Why was this not widely reported? If this is the official position of the
Bluetooth group (and not a prank) then they should be ashamed of it.
(And someone who has a copy of the list and isn't party to the quoted
"license agreement" should mirror it.)
Walter
Bluetooth vs. BlueZ project
Posted Dec 15, 2005 11:10 UTC (Thu) by wookey (subscriber, #5501) [Link]
Indeed, and especially as the reason given appears to make no sense (as described on the page the artivcle linked) I don't understand why Mr Holtmann didn't just tell them to get lost. But still, if he doesn't feel able to be bolshy for whatever reason, then clearly somone else should put the list up. This sort of thing is too useful to be lost to the craziness of the BQA.
And whilst I am here another really useful hardware list is this one:
http://ralink.rapla.net/ (all the ralink-based devices for USB, PCI and PCMCIA and as ralink are FS-friendly they all work).
It is a real pain when you want to buy some new hardware (I'm currently looking for a DVB-T receiver) and you have to spend _ages_ on the net finding out what is and isn't supported. A comprehensive database would be a huge boon.
Bluetooth vs. BlueZ project - internet archive to the rescuee
Posted Dec 17, 2005 16:36 UTC (Sat) by nealmcb (subscriber, #20740) [Link]
Here is the 2004-11-14 version:
http://web.archive.org/web/20041114015211/http://www.holt...
Ralink - not quite supported
Posted Dec 18, 2005 0:49 UTC (Sun) by midg3t (subscriber, #30998) [Link]
Be careful with Ralink-chipped devices. I bought one and am still waiting for a driver that works on my big-endian laptop.
Ralink - not quite supported
Posted Dec 21, 2005 8:38 UTC (Wed) by wilck (subscriber, #29844) [Link]
Another ralink caveat: the user-level interface is different than for most other wireless devices. For example, WPA is handled through wpa_supplicant for most wireless cards, but not for ralink.
This is unfortunate, because distos rely on a homogeneous user-level interface with their configuration tools. Getting WPA to work with ralink card and SuSE 9.2 required a fair amount of hacking, while it would have been easy with any card supported by wpa_supplicant.
Ralink - not quite supported
Posted Dec 26, 2005 15:47 UTC (Mon) by cortana (subscriber, #24596) [Link]
These criticisms only apply to the old rt2400/rt2500 drivers that developed from Ralink's original driver release. The maintaners are currently working on rt2x00, a rewrite of the driver that uses the new unified iee80211 stack, and so will hopefully end up in the Linux kernel itself.
The project released its third test release last week; if you have a Ralink device and some time to spare, please try and see if you can get the driver to associate with your access point. Details at http://rt2x00.serialmonkey.com/phpBB2/viewtopic.php?t=693
"Just works with Linux"
Posted Dec 15, 2005 11:43 UTC (Thu) by migpc (subscriber, #24484) [Link]
There is a effort going on in this front called compatiblelinux:http://www.compatiblelinux.org
It's been launched by Junta de Extremadura known for it's gnuLinex (a.k.a. Linex) and an Spanish university called Rey Juan Carlos.
Its hardware database is not really extensive but pretty useful nevertheless.
Just my 2 cents.
M*
"Just works with Linux"
Posted Dec 15, 2005 12:39 UTC (Thu) by zooko (subscriber, #2589) [Link]
I find myself less and less interested in Linux, and more and more interested in Free Software. I think part of this is because of the success of Linux -- it works reliably enough that I can stop tinkering with it and just use it as an invisible underlying platform on which to build.
I'm kind of hoping that a lot of LWN subscribers and writers feel the same and that LWN will gradually transition into the FSWN or OSSWN. I work with Linux, Windows XP, and Mac OS X every day (except no Windows XP on weekends). Someday I might jump from Linux to Solaris at some point, but hopefully 90% of the LWN content would still be relevant and interesting to me after the switch.
So, to swerve back onto topic, I too am very disappointed when I find out that some hardware has been advertised as "it works with Linux" and then it turns out that it requires proprietary drivers.
So rather than a "It Just Works With Linux" database, I would like to see a "It has freely available, open specs and/or Free Software drivers" database. For example, the OpenBSD folks have put some serious effort into opening up wirelss cards. Where is the summary of the current status of that effort?
"Just works with Linux"
Posted Dec 15, 2005 14:03 UTC (Thu) by bronson (subscriber, #4806) [Link]
I couldn't disagree more. LWN is valuable to me precisely because it covers a single subject and covers it extremely well. If you want a Solaris periodical, there are at least five to choose from and most are free to qualified businesses. Despite the CDDL, Solaris is still a closed platform. Why on earth would you want to find any relevant content in LWN (much less 90%?!)
A few years ago, with the exception of Robert Love and GregKH, LinuxJournal articles started getting very shallow and broad. Just this year I reluctantly let my 8-year susbscripiton lapse. Back in the day it was a really great magazine; now they could just call it LDAP Installation Monthly. But, in theory, it appeals to a broader audicence.
"Just works with Linux"
Posted Dec 15, 2005 14:42 UTC (Thu) by zooko (subscriber, #2589) [Link]
You wrote: "LWN is valuable to me precisely because it covers a single subject and covers it extremely well."
I agree that LWN has usually high-quality content. But is it narrow in focus? With the exception of the "Kernel" page, almost all of the content is interesting to our hypothetical reader who doesn't use Linux but who uses Free/Open Source Software. Look at the contents of this current edition:
Front Page:
* "Just works with Linux". Well, that's how this conversation got started. corbet (and I) feel that the important and interesting issue is more general than Linux -- it's Free Software drivers and open specs.
* GStreamer to support DRM. Not Linux-specific.
* Gnome v. KDE. Not Linux-specific.
Security:
* Community Help as an Attack Vector. Not Linux-specific.
* New vulnerabilities. Approximately 25 out of 27 vulnerabilities are not Linux-specific.
Kernel:
All Linux-specific
Distributions:
All Linux, except Nexenta (Solaris) and PC-BSD (FreeBSD).
Development:
By my count something like 35 out of 38 items are not Linux-specific.
Linux in the news:
Honestly I don't read this page, but it's probably fairly Linux-specific.
Announcements:
I don't read this page, but I'd guess around 50% of announcements are Linux-specific.
Letters:
I don't read this page, but I'd guess around 50% of letters are Linux-specific.
So in sum, the stuff that I personally really like in LWN is:
Front page, editorials (occasionally Linux-specific)
Development page (usually not Linux-specific)
Kernel page (always Linux-specific)
The stuff that I personally mostly like in LWN is:
Distribution page (almost all Linux-specific)
The stuff that I personally barely like in LWN is:
Security page (usually not Linux-specific)
The stuff that I personally don't read in LWN is:
Linux in the News, Announcements, Letters
You mentioned that if I were interested in other things that Linux that I could read other periodicals, but in fact I am not aware of any periodical devoted to Free Software which has the regular high-quality (and not-Linux-specific) content that LWN does. If there is one, please let me know!
LinuxJournal
Posted Dec 16, 2005 18:47 UTC (Fri) by pflugstad (subscriber, #224) [Link]
I couldn't agree more - LJ has gone downhill a lot. The issue that had the "supporting high-end hardware" (June 2005) trumpted on the front and the article proceeded to have almost zero "high-end" content (it was an article about RHEL4) was my last straw. I let my many year (I had issue #10 IIRC) subscription lapse...
LinuxJournal
Posted Dec 17, 2005 18:37 UTC (Sat) by N0NB (subscriber, #3407) [Link]
Last year I took part in a six month readers survey for LJ. At the end I took an opportunity to add a number of personal thoughts. I've been a subscriber since late '96 and I watched LJ grow from mostly a hobbyist centered publication to one focused on the sysadmin in a large organization to the past few years focusing on LAMP. I asked for more home/hobbyist articles and I've noticed a few more over the past year.
I would like to see the direction Nick Petreley takes LJ before deciding the fate of my subscription.
OpenBSD's efforts
Posted Dec 15, 2005 17:36 UTC (Thu) by rfunk (subscriber, #4054) [Link]
You can find the results of OpenBSD's efforts on their hardware pages (e.g. the one for i386).When I recently bought a wireless card this was quite useful to me, in combination with Jean Tourrilhes's Linux wireless drivers page.
I think that OpenBSD's efforts to open up hardware have been extremely beneficial to the Linux community, and I hope they continue.
What about vendors?
Posted Dec 15, 2005 13:38 UTC (Thu) by tclark (guest, #32451) [Link]
It seems to me that we're letting hardware vendors off the hook here. The people who produce and sell the hardware are in a good position to evaluate its compatibility with Linux, and they have a clear interest in telling us about it.
What we need is a web site where hardware vendors can register their Linux compatible hardware, certifying that it meets clear and strict compatibility standards. Perhaps certified hardware vendors could display a "just works with Linux" seal on the product.
We would rely on vendors to be truthful, but we could keep them honest by requiring that vendors supply sample hardware for testing on demand and periodically spot checking. Vendors failing such a test could lose their certifications for a year or so,
What about vendors?
Posted Dec 15, 2005 14:01 UTC (Thu) by droundy (subscriber, #4559) [Link]
Indeed, I think the key is to get some sort of "works with vanilla linux" certification program, andallow vendors (and retailers) to specify that they're certified. Once vendors realize that being on
this list is a way to attract customers (even if it's just a few), then as long as they already qualify
it'll be well worth their while to keep the database up to date.
If the certification requires that the device work with vanilla linux, then it'll also encourage
vendors to submit patches to the mainline kernel. I suppose I'd also like a "sort of works with
vanilla linux" to handle devices that can work with reduced functionality on the vanilla kernel
(e.g. no 3D). Often, reduced functionality is still plenty--I can't remember the last time I ran a
3D program...
David
What about vendors?
Posted Dec 15, 2005 14:21 UTC (Thu) by bronson (subscriber, #4806) [Link]
Margins on consumer hardware are typically so razor thin that the vendors just don't have the means to support Linux. Desktop Linux users make up only a tiny fraction of the potential market yet Linux support tends to be an utter nightmare (what kernel version are you running, what distro, what patches, ...)
Until a noticeable fraction of CompUSA customers are running Linux desktops, hardware vendors aren't on the hook for anything.
What about vendors?
Posted Dec 15, 2005 15:13 UTC (Thu) by pizza (subscriber, #46) [Link]
That's just it -- If they provide documentation to interested developers, or provide a full open-source driver they *don't* have to provide anything more in the way of support.
Most of the "support nightmare" of Desktop Linux goes away when your driver is provided as full source, and goes away almost entirely if it goes into the kernel -- they only need to actually support the developers who are responsible for maintaining the driver.
And it's alwo worth mentioning that those selling "consumer hardware" with "razor thin" margins aren't actually developing their own drivers for Windows either; by and large they're handed to them from their ODMs or the chipset manufacturers. The latter is especially true of wireless cards. The OEMs would love to hand out Linux drivers, but they don't have anything to give to their customers because their upstream provider won't supply it, or let them.
What about vendors?
Posted Dec 15, 2005 17:32 UTC (Thu) by tzafrir (subscriber, #11501) [Link]
I don't want them to support "Linux". I want each vendor to support at least one target Linux distribution (be it a custom one), that should be free enough.
Specifically: no binary drivers allowed. Not even through wrappers such as ndiswrappers.
If the product works well with ReferenceLinux, chance are other linux packagers will be able to adapt it to their distro. And if not: the community will.
HW/SW Linux certification
Posted Dec 15, 2005 15:07 UTC (Thu) by bkw1a (subscriber, #4101) [Link]
I like the idea of a "just works with Linux" sticker program, givingvendors the right to put a recognizable logo on their packaging if
they can verify that their hardware works with the mainline linux kernel.
(Of course, there's still the question of whether it'll work with
the available user-space tools in a given distribution, but at least
it's a start.)
I'd also like to see a similar program for software. More and more,
I find that software that claims to "work with Linux" really only
works on some very recent distributions. I support ~ 200 Linux
machines, and my users frequently complain that they've tried to
install some software, only to find that we'd need to upgrade to a new
distribution version to get the software to work.
This is only partially the fault of the software vendors. They
understandably can't compile many different versions of their
software for many different distribution versions.
The second big contributor to the problem is the lack of support for old
APIs. Under Windows, I can install the current version of Acrobat
reader on machines running Windows 95,98,NT,2000,XP or 2003. Software
written today will still run on Windows versions 10 years old. In
contrast, in the Linux world, a QT- or gtk-based program written
today won't even run under a distribution released last year.
This is frustrating for both vendors and users.
I don't know the solution to this second problem. Maybe just
more-stable APIs. Maybe some way of gracefully failing when a
newer feature is found not to be available. I think it's
a problem that really needs to be addressed soon, though, if
Linux is going to be accepted widely on the desktop.
FUD, FUD, FUD
Posted Dec 15, 2005 21:52 UTC (Thu) by khim (subscriber, #9252) [Link]
Under Windows, I can install the current version of Acrobat reader on machines running Windows 95,98,NT,2000,XP or 2003. Software written today will still run on Windows versions 10 years old.
Check your facts - or you'll look ridiculous.
Let's check your facts, shell we ?. I'm happy "10 years old Windows" user. 10 years old Windows is "Windows NT 4.0" or "Windows 95". Let's start !
I'd like to install Acrobat reader. Latest and greatest is "Acrobat reader 7.0.5". Can I use this ? Of course not! Acrobat reader 7.0.5 is only compatible with Windows XP - not even Windows 2000 is enough (not without SP2 anyway). Oh, too bad. Can I use Acrobat reader 6.0.1 instead ? Nope: you need Windows 2000 or Windows 98SE (Windows 95 or Windows 98 are not supported). Oooh, ok. How about Acrobat reader 5.0.5 ? Still no cigar: Windows 95 or Windows 98 is required (Windows95 user is happy at this point: he/she got only two generations old Acrobat reader; Windows NT 4.0 user is still not). Argh. Can I at least use Acrobat Reader 4.0.5 ? No: even this thing is only compatible with Windows NT 4.0 SP3, not my "10 years old Windows". My only hope is ancient Acrobat Reader 3.0.1... Four generations old...
How is this for "support for old APIs" ? Sure: I can install ancient Acrobat Reader 3.0.1 on my Windows XP SP2 - but that's not such a big achievment: most distrost do include compatibility packages.
P.S. Big thnx for selection of package - some new programs are still runnable on old versions of Windows. Acrobat reader is not one of them.
HW/SW Linux certification
Posted Dec 16, 2005 2:38 UTC (Fri) by zblaxell (subscriber, #26385) [Link]
I seem to keep running into the opposite problem: Someone asks me to install a Linux binary they found somewhere (in some cases after they paid six figures for it), and I hear that the thing supposedly only runs properly on six-year-old versions of Red Hat. Often the word "run" is generous, and in a number of cases I have forced vendors to admit that their software limps like a wounded waterfowl at best on any Linux version, but there is no hope at all of this software working on a modern Linux distro.
A quick phone call to the vendor followed by some not-so-quick escalation up the support ladder and I end up talking to some engineer who is working on porting the application to a three-year-old version of Red Hat. It should be ready for general release next quarter.
Fast forward a year, and the sales weasel comes to install their "new" software on the now-four-year-old version of Red Hat. I thoughtfully dig through my CD-R archive and provide a machine with exactly the specific old Red Hat version requested. I even provide a second machine with all the official updates to that Red Hat version applied. Both are running a vanilla configuration unless the weasel specified otherwise. Neither works out of the box with the weasel's software, although they fail in different ways.
These days I'm sick of playing the "supported platform" game. I just tell the weasel what we have on our desktops ("here's the install CD guys, have fun"), and that we'll pay only for whatever parts of their software work after they are installed.
HW/SW Linux certification
Posted Dec 16, 2005 14:42 UTC (Fri) by khim (subscriber, #9252) [Link]
It's the same in Windows world. I know one case where vendor two years ago (that's 2003, folks!) finally started to support... what ? Windows 2003 ? Windows XP ? Windows 2000 ? No! Windows NT 4.0SP6a!!! Yup - till that time they only supported "Windows NT 4.0SP5"! Situation is quite similar to Linux versions, right ?
HW/SW Linux certification
Posted Dec 16, 2005 23:08 UTC (Fri) by zblaxell (subscriber, #26385) [Link]
I know of one vendor who is *at this moment* in the process of porting their application from Windows 3.1 (with 16-bit DLLs and everything) to Windows 2000.
On the other hand, at this moment the old Windows 3.1 binaries do work on Windows 2000 if you install all the relevant backward DLLs...I wish the major Linux distributions made it easier to do that.
SW Linux certification
Posted Dec 22, 2005 22:07 UTC (Thu) by anton (guest, #25547) [Link]
>a QT- or gtk-based program written>today won't even run under a distribution released last year.
>This is frustrating for both vendors and users.
>
>I don't know the solution to this second problem.
Possible solutions:
- Build on old distributions. The resulting binaries often work with
new libraries. However, sometimes they don't, but that can be
tested relatively easily (test it on a system with new libraries).
The disadvantage is that the program would have to restrict itself
to the ancient API.
- Use static linking. This has some disadvantages wrt security
updates and on-disk and in-memory size, but it could be supplied as
a fall-back option (the old system probably does not have the
updates anyway).
- Compile the program from source (or use a source-based distro like
Gentoo). At least that gets rid of some gratuitious
incompatibilities thrown in by linking with modern libraries,
although not from the API dependences.
What about vendors?
Posted Dec 15, 2005 15:55 UTC (Thu) by mmarsh (subscriber, #17029) [Link]
What we *need* is to be very clear with vendors about our requirements. If you're buying a Linux-installed system, and you want only open-source drivers, you need to tell the vendor exactly that. Make it crystal clear. Tell them, "If, when I receive this system, I determine that any of the hardware is using proprietary or other than non-free drivers, I will return the system and demand my money back." They can make all the claims they like (well, not *all*), and if they ship something that's not completely free, they can excuse themselves by citing the lack of technical knowledge on the part of the salesperson. Vendors will get away with this for as long as we let them.
What about vendors?
Posted Dec 15, 2005 16:55 UTC (Thu) by tlw (subscriber, #31237) [Link]
Agreed. It's too easy for a vendor to write "works with Linux", but that's not specific enough. "Works with Linux" should be defined to mean:
- specifications are available
- driver is open-source
- driver is integrated into the kernel.org kernel
It's happened to me before where I'll obtain a device, it comes with a CD with source code (great!)... for kernel 2.2.20 and 2.4.9. It would be nice if "works with Linux" were a shorthand form of saying "we work with the Linux community".
What about vendors?
Posted Dec 19, 2005 7:28 UTC (Mon) by rakoenig (subscriber, #29855) [Link]
"driver is integrated into the kernel.org kernel"
Ok. I'm working at a vendor and my job is to make sure that Linux is running on our PCs.
Your statement is nice, but I can tell you what we're currently going through:
Launched a PC with a new southbridge. That caused problems because it wasn't supported in the kernel, so Linux was unable to "see" the SATA drives. We did a lot of effort to fix this issue and since 2.6.14 the driver is in the vanilla kernel.
But that actually doesn't have much impact on our customers: All the current distributions (RHEL, SLES, Debian, Fedora Core 4, SUSE Linux 1O...) are shipped with a kernel that is less than 2.6.14. The result is that if a customer tries to install he runs into the problem that his installation program will report that there are no hard drives!
So fixing issues on kernel.org kernels is just the half work. The other half must be done by the distributors. And since this is an issue that affects installation it leads to the need of new installation media, that makes the issue more complicated. It would be easy to do a kernel update online, but as long as you're unable to create a running system from your
installation media you're probably not happy.
Well, for that issue there is light on the horizon: Fedora Core 5 Test 1 has the fix, so is with openSUSE 10.1 alpha. SLES 10 will ship with it and Red Hat promised to have the patch in Update 3 of RHEL4. So soon the situation will get much better for the customers.
But "driver is integrated into the kernel.org kernel" is not enough.
One solution could be hardware certification, but unfortunately there are not much certification programs around. And a RHEL4 certification won't imply that the machine is running fine with Debian/Sarge.
What about vendors?
Posted Dec 19, 2005 14:58 UTC (Mon) by dlang (subscriber, #313) [Link]
it's true that getting the hardware to work with the kernel.org kernel doesn't instantyly make it work with all distros, however once or twice a year (sometimes more) the distros update their kernel from kernel.org. (this is one of the reasons for the 2.6 development model, to allow the distros to sync up more frequently and reduce the lag (and divergance) between the different kernels.
What about vendors?
Posted Dec 21, 2005 9:26 UTC (Wed) by wilck (subscriber, #29844) [Link]
once or twice a year (sometimes more) the distros update their kernel from kernel.org
This isn't true for the enterprise distros. They stick to the same base kernel for years, just backporting new drivers as they see need.
What about vendors?
Posted Dec 20, 2005 16:11 UTC (Tue) by kalou_ti (guest, #32988) [Link]
"Perhaps certified hardware vendors could display a "just works with Linux" seal on the product."
When I went to England about 2 years, I was astonished and happy to see such a sticker on many hardare. YES, I saw hardware sold with a Pinguin to confirm that the hardware runs Linux.
I have always been wondering how this has been done in the UK, and would be happy to have feedback on this from someone living in the UK out there.
What about vendors?
Posted Dec 21, 2005 9:23 UTC (Wed) by wilck (subscriber, #29844) [Link]
What we need is a web site where hardware vendors can register their Linux compatible hardware, certifying that it meets clear and strict compatibility standards.
Who is going to run that web page and be the certification authority? And who is going to pay for it? The vendors of, say, USB modems or flash card readers??
Certification is an expensive task and I fear most HW vendors don't think it'll pay off (with the notable exception of expensive server hardware, such as high-end RAID controllers).
No antique hardware please
Posted Dec 15, 2005 19:46 UTC (Thu) by jenro (subscriber, #7024) [Link]
Whatever solution is found for this hareware-datebase, there is one pointthat should be kept in mind: if this list should be useful for bying new
hardware, this hardware should still be for sale.
For example the FAQ of the ISDN4linux-project lists a lot of working
hardware. But you can't buy most of these things today. So if you want to
buy a new ISDN-card, and you neither want to spend lots of money for an
active card, nor use a proprietery driver, you find: nothing.
No antique hardware please
Posted Dec 22, 2005 19:10 UTC (Thu) by barrygould (guest, #4774) [Link]
I disagree. If it's still available on eBay, it should still be listed.
Example...
Prism II / 2.5 network cards are scarce in retail, but commonly available on eBay.
(And they are the only cards that work with some wireless security tools.)
Barry
"Just works with Linux"
Posted Dec 15, 2005 20:43 UTC (Thu) by stephenjudd (subscriber, #3227) [Link]
I feel I should commend Dick Smith Electronics here in New Zealand, who sell a lot of "house brand" components and peripherals. They put great effort into ensuring Linux compatibility, and most important, they put a nice Tux logo on the box next to the Windows and OSX logos. So just browsing the shelves, you know that if it has the logo on it will work when you get it home.
"Just works with Linux"
Posted Dec 15, 2005 21:57 UTC (Thu) by khim (subscriber, #9252) [Link]
This is great. What is not so great is to buy stuff with Tux on side, bring it home and... read instruction for NDISWrapper installation! We need some certification program. At least then Tux logo will mean something...
"Just works with Linux"
Posted Dec 26, 2005 23:57 UTC (Mon) by csawtell (subscriber, #986) [Link]
> So just browsing the shelves, you know that if it has the logo> on it will work when you get it home.
No you don't, what that logo means in the DSE case is that an experienced
technician said that he got it to go with some unmentioned kernel version,
on some unmentioned Linux distribution at some unmentioned time in the
past. Whether or not that, or indeed the very limited documentation
offered, has any relevance whatsoever to your current situation is pure
coincidence.
I was able to get one piece of kit ( a USB <=> RS-232 adaptor ) to go
after hours of futzing around with the Google, the actual chip
manufacturers web pages, and a kernel re-compile. The cryptic
documentation in the booklet was so out of date that it was totally
useless.
The other ( a USB sound system ) was a _complete_ failure. DSE, to give
them due credit, did give us a hassle-free refund.
I'm happy to go on record saying that in my opinion DSE need to get their
Linux support act together somewhat more effectively than it is now before
they use the ( licensed ? ) Linux Penguin logo on their product packaging.
The role of OSDL
Posted Dec 15, 2005 22:33 UTC (Thu) by bojan (subscriber, #14302) [Link]
> Perhaps an industry group (OSDL, say) would like to serve the community by taking this task on.
Absolutely. They employ people to write Linux, their members are hardware manufacturers and distribution makers. Basically, people with a lot of interest in getting Linux to "just work".
What they need to do is maintain a matrix of hardware that just works against vanilla kernel. What they need to allow their distributor members to do is to maintain a matrix against their own distribution (RHEL, SUSE, Debian etc.). Distributors should deprecate their own hardware compatibility lists in favour of this one. Hardware manufacturers can then work both with OSDL and distributors on the maintenance of the matrices.
Each entry should contain the level of "support" (i.e. what does and doesn't work for that hardware component/driver). People should be able to find the entries by multiple criteria, such as driver name within the kernel, chipset used on the hardware component, manufacturer name/code of the component, functions of the component etc.
All support that is based on binary drivers should be clearly marked with warning signs that link to an explaination as to why it is imporant that hardware manufacturers release their drivers under a kernel compatible licence (that kind of thing should not be a surprise at _Open_ _Source_ Development Lab). We read some good explainations recently here at LWN, prepared by kernel maintainers themselves, so that shouldn't be hard to do.
Big task by all means, but when done by so many organisations and individuals in one location, it shouldn't be impossible.
Oracle's 10g Release 2 (10.2) for Linux x86-64
Posted Dec 16, 2005 16:39 UTC (Fri) by dwkunkel (subscriber, #5999) [Link]
Oracle's sofware requirements for installing the 10G database for x86-64 Linux provide an example of what is needed from vendors.
"Just works with Linux"
Posted Dec 26, 2005 17:54 UTC (Mon) by tobler (guest, #34758) [Link]
Where is freedom to speech and publish information?
When I am going to buy something computer equipment, I'd like to know in advance what works with my system. If I can't find, then I just buy something and ask if I can return it. So computer shops have to start finding compatibility information - or I just return equipment and ask my money back.