Slashdot Mirror
IBM Releases AFS
Raleel writes: "IBM has released the source code to AFS for AIX 4.2, Digital/Compaq UNIX 4.0, Red Hat Linux 6.2, Solaris 2.6 and 2.7, and Windows NT 4.0. You can download it from here. It is under IBM's Open Source license." This was supposed to be released a while ago, but it's good to see IBM following-thru. For more information, see our article regarding the open sourcing of AFS and the article from 1998 regarding the porting effort.
Once AFS is truly considered dead and no longer supported at all... there will still be some legacy systems around using it. Open source could possibly be the only avenue of support for these systems in the future.
sorry, my bad. It is free, but incompatible with the GPL.
http://www.gnu.org/philosophy/li cen se-list.html
---
Desperation is a stinky cologne
Please note folks... AFS is not JFS. AFS is a remote file-sharing protocol, like NFS, which fails to suck in most of the ways that NFS sucks.
They are related in that they were both created at CMU as part of the Andrew project. The applications you are thinking of are probably part of the Andrew User Interface System. It's a suite of integrated applications for Unix. Check the web site for more info.
--Sam
ezmail is an AMS client; the release of the AFS source code does not impact the status of ezmail.
--Sam
> So... They're going to release it, and not support it. Joy...
Gee, next thing you know, IBM is going to stop supporting Linux, Emacs, gcc, flex, bison, and gmake, too!
Man, you're thick. Open Source means that they're releasing the source code in some free manner. It does not mean that they're opening their wallet.
If the authors of free software had to support it out of their own pockets, how much free software do you think there would be? How many people would actually bother to write any?
Imagine the world without it..
"Hello, GNU Software. This is Richard M. Stallman speaking."
"Yes, sir. I know you're having trouble with Emacs, but ever since Sun and Amdahl pulled their
code out, it hasn't been very stable under that platform. I think it's a bug with your vendor's crappy C compiler or their lexer, but since there is no other alternative cc or lex, I can't really narrow it down very well. I'm also really too busy manning the phones to code much these days...
"No, sir, there aren't any other support people here -- nobody else wants to work for free. I'm afraid you'll have to deal with me."
"Whoah! I'm sorry sir -- I have to go! Microsoft SourceSafe just ate my Emacs v5 source code -- boy I wish there was some sort of a revision control, or concurrent versioning system available for UNIX!"
[click]
>IBM - "We're releasing this, isn't this great?!"
>
>YOU- "wow, this is awesome, so how does bla bla bla..."
>
>IBM - "RTFM!"
It must suck to be illiterate. You have my condolences.
Personally, I'd much rather have a piece of code that I had to RTFS or RTFM to use rather than writing it from scratch myself. But then again, you're probably one of those "Gimme, Gimme!" Americans.
--
Do daemons dream of electric sleep()?
Sadly they are not selling support. I hope that one day they will see that is the best thing for Open AFS.
------
My opinions are mine alone and do not represent that of my employer.
# vos move user.janeane afs-1.iastate.edu /vicepa afs-7.iastate.edu /vicepe
That's me moving my wife's homedir volume from one server to another while she's using it AND she never even notices it -- everything keeps working perfectly even while it is in transit.
That rocks.
Shut up, be happy. The conveniences you demanded are now mandatory. -- Jello Biafra
IBM bought Transarc about 5 or 6 years ago when all the pundits and industry gurus though DCE was going to be the next big thing.
There product line was the file server, a transaction monitor (ENCINA) and some extras like transactional file and queue systems. The best selling product was a UNIX version of the mainframe transaction monitor CICS, which was remarkted by IBM as CICS/6000.
DCE never really happended, ENCINA never really took off, and, CICS sales slowed down. So it wasn't a brilliant buy.
The Andrew File System though is pretty good apert from the fact that it depends on DCE, it is certainly much better than NFS in all respects.
I think as far as IBM is concerned its a case of we can't sell it so lets give it away.
Old COBOL programmers never die. They just code in C.
Have you tried modifying their makefile? I just ran into the same problem and my first thought was to just *try* it for 2.2.17 by adding it in.
I absolutley agree that the documentation is wonderful. You don't need support except for bugs with the thorough doc. I am not sure about the "Even if AFS doesn't DO anything". But, overall you are very correct.
-------
My opinions are mine alone and do not represent that of my employer.
The Andrew File System though is pretty good apert from the fact that it depends on DCE, it is certainly much better than NFS in all respects.
AFS does not depend on DCE - it depends on a hacked version of kerberos IV.
It is better than NFS in most respects, but not all:
Wangden
an AFS sysadmin for 2+ years
-- "You're not fooling me, young man - it's turtles all the way down!"
While Arla works great for some situations (I use it all the time when doing standard development stuff), it has one huge drawback: It doesn't do chunck caching. Transarc's client can cache a chunk of a file, rather than the entire file. Arla can only cache the entire file. Normally not a big deal. But when you start working with huge data sets, you can really get hosed if your server is a little slow or far away. That being said, I really like the fact that ARLA can actually be stopped without rebooting. It always seems that trying to stop transarc's client causes a kernel panic.
Keep an eye on www.openafs.org. This is where the real action will be happening on this project, not at the ibm.com site.
Wangden
-- "You're not fooling me, young man - it's turtles all the way down!"
SGI just released the DMAPI implementation for their XFS filesystem, and Sistina has DMAPI on their todolist for GFS.
In addition, there are the openxdsm project, and mfs (a stackable filesystem that add HSM [not DMAPI] to a regular filesystem). I believe Unitree have their own DMAPI capable filesystem ported to linux, and they have VFS modifications to allow DMAPI (like openxdsm) on any linux filesystem in the pipeline. Unfortenately neither are open source.
I was the DCE/DFS Administrator / Architect at Brown University. DFS/AFS is very scalable, a monster to administer, but I didn't complain, I just rolled my own Perl scripts to administer the beast. I figured as long as this is a beast to administer, the pay will be great! Anyway I wanted to put a AFS server online for friends / collegues to access via my cable modem. But... I hit a compile snag at afsmonitor.. it looks like problems with some of the supplied header files. I get 75% compiled. Does any one else run into this? I'm hoping to have a version compiled for LinuxPPC soon! And a fully configured AFS cell. I am also going to work on a AFS/DFS gateway
I really would like to have an AFS Cell up on my home LAN and accessable via my cable modem, also would like to experiment and set up gateways to other hobbyist cells! Anyone interested! Yes AFS/DFS is very powerfull, and there are plenty of ways to manage it. I used my own perl scripts. Now I tried to compile the code that was on IBM's site. Here the error I get: ../afs/afs_analyze.c: In function `afs_Analyze': ../afs/afs_analyze.c:334: storage size of `opStartTime' isn't known ../afs/afs_analyze.c:334: storage size of `opStopTime' isn't known ../afs/afs_analyze.c:334: storage size of `elapsedTime' isn't known
make[4]: *** [afs_analyze.o] Error 1
make[4]: Leaving directory `/home/jerryn/afs/i386_linux22/obj/libafs/MODLOAD- 2.2.5-15-MP'
make[3]: *** [linux_compdirs] Error 2
make[3]: Leaving directory `/home/jerryn/afs/i386_linux22/obj/libafs'
make[2]: *** [libafs] Error 2
make[2]: Leaving directory `/home/jerryn/afs'
make[1]: *** [install] Error 2
make[1]: Leaving directory `/home/jerryn/afs'
make: *** [all] Error 2
Did anyone else run into this?
I am using kernel 2.2.14, and yes I have all configured ok! And most of the code compiles.
just afs_analyze fails.
my email address is jerry-normandin@usa.net
I think we should start a newsgroup on egroups
for all of us who want to create a massive WWW
AFS implementation. I've got bandwidth from my
cable modem, and I can limit the number of concurrent connections to my AFS server with my
firewall.. All I need to do is get all the code compiled and I can easily configure this puppy since I used to be a DCE/DFS admin!
Anyone interested?
I did a symbolic link of /usr/src/linux-2.4-test10 /usr/src/linux-2.2.14, but it wasn't fooled. It checked the version.h header in the kernel source to verify the correct kernel version. Maybe if I edit the version header manually to 2.2.14...hmmmmmm
"Recursive bipartite matching"- try it!
OK, call that subject the reaction to reading too much FUD on here lately, and at having that FUD directed at a License *I* release code under; my appologies.
That said, the OSD, is derived from the DFSG. Both of which clearly contain points related to free (as in speech) code such as #3, #4, #6, #8 and #9. As well as the expected free (as in beer) points you're thinking of: #1, #2, #4 - 7, and #9. As I said, if they don't meet your definitions of free (as in beer and speech) then WHAT DOES?
Now, IANAL, however I have read through both licenses in detail... what have I missed? What does your obviously superior legal knowledge expose that makes the IBMPL not free as in beer AND free as in speech? And what is it about the IBMPL that prevents software released under it from being included in a Linux distro? are you somehow indicating that just because the kernel and much of the runtime is released under the GPL that this means ALL software in the distro must also be GPL? If that's what you're thinking then you've missed the point of #9.
Yeah, and they had to go through the code and remove all the names and initials of all the past developers. For some reason.
One of the things nice about what happened here is that a slew of documentation was released with the software - in general, I have noticed relatively sparse documentation around new open software. Not a complaint, just something I noticed.
No, it doesn't - it says that if you pay for 'IBM AFS' support you won't get 'Open AFS' support - they're distinguishing between products (you might not be able to get 'Open AFS' support either, I don't know, but the FAQ doesn't preclude it).
--
...or am I missing something?
I was wrong. Sorry about that - maybe I should learn to read properly or something.
[watching karma drop as previous post gets modded down...]
--
...or am I missing something?
AFS looks very cool. I've looked around at the FAQ and done some searching but I haven't been able to determine whether it supports SSL for transporting the raw data. However, I have observed that it supports secure authentication. Does AFS provide the capability to encrypt all data traffic with SSL? If not, might we have some luck including that feature now that the package has been open-sourced?
So on the one hand we should make sure they get lots of positive attention to encourage further OS releases by companies (otherwise as you say 'they won't bother'). Yet also stop patting them on the back? Well you can't have it both ways! If _you_ don't like it, then don't use it. And if no one uses it, big deal, it's not costing you anything at all. Just be glad that the effort is being made, even if it is only with something _you_ don't rate.
Rant: off
--
Azrael - The Angel of Death
I don't get it. I suppose some people just like to piss and moan about how their asses aren't being correctly kissed on BOTH cheeks. WTF could there possibly be to complain about this announcement. AFS is great A robust distributed file system that was designed to handle huge gobs of data and huge numbers of users. And this is bad how?
If you think this is something to complain about then shut the fuck up and work on the code.
Previously, I was constantly stuffing my face with packets, unable to stop this neurotic activity because of demands on the job.
AFS is great, I lost 500KB bandwidth _per second_.
Yours,
Mr. Switch
-- Bird in the Bush: The Renewable Energy Blog http://www.birdinthebush.org
Everyone who knows Martin knows that he's a schizophrenic crack-addict. No right-minded person would desire a return to EZMAIL.
I'm using Arla now because it compiles on 2.4, but arla isn't terribly stable yet even as a client, and I've heard that it's even worse as a server.
Too bad ARLA won't be able to use code from OpenAFS...
---
pb Reply or e-mail; don't vaguely moderate.
pb Reply or e-mail; don't vaguely moderate.
IIRC, Coda was originally under a BSD-style license which was changed to GPL 6-12 months after they started work.
1) how the hell did this get mod'd up?
...I guess this proves that just about anyone with make and gcc thinks they need to be compiling every source tarball they miander across while aimlessly wondering the 'Net.
2) install either a tarball or packaged version of the kernel source code, and chances are your problems will all disappear!
What would you even do with ASF? If you couldn't figure out how to compile the code, chances are you were going to be hopelessly doomed at using it!
-C
"This above all, to thine own self be true"
- AFS is relatively bug free and stable, while CODA is still an alpha/beta quality product that has plenty of disclaimers about losing data.
- AFS is well documented and somewhat easy to use, while I personally had a tough time figuring out anything whatsoever about CODA.
-AFS supports multiple "cells", while last I checked CODA supported one.
- I think AFS is generally targetted at large installations (for example nearly all of IBM uses it) while CODA is intended for smaller ones.
- CODA supports "disconnected operation", while AFS dies even if you just have a temporary network glitch.
If any of this is now inaccurate, please feel free to correct me, or mod me down to flamebait.
STFU about slashdot bias.
This was posted to info-afs recently (subscritpion information at http://www.transarc.ibm.com/Support/common/mailing ListRequest.html , archive at http://www.mail-archive.com/info-afs@transarc.com/ )
From: Derrick J Brashear
To: info-afs@transarc.com
Subject: OpenAFS lists, cvs to be available at openafs.org
As soon as the relevant DNS changes happen, lists devoted to openafs development will be available at openafs.org, and a cvs archive will also be available. A preview of the site is available at www-openafs.central.org.
-D
AFS ACLs allow users to come up with quick fixes to permission problems, and they ultimately leave a complete and often unsecure mess behind, something that the system administrator has to deal with.
The tedious but simple NFS/UNIX permission schemes make it much easier to figure out what is going on, and they force people to think carefully about what kinds of groups they want, rather than creating them willy-nilly by accident.
That's only one of the many problems with AFS in my experience. That isn't to say that AFS is all bad--it works reasonably well in its original environment. But it isn't the universal answer to .com's or corporate computing.
I think it's fine AFS has been released by IBM. I also think the open source and Linux community should think carefully about whether to invest any resources in it. There are probably better directions to go in and better projects to spend time and resources on.
Yet more funky stuff enters the OSS world... The question now is 'What will Big Blue opensource next?', and who wants to port it? :o)
ManicHawk - Just because you're manic doesn't mean the walls aren't bouncy
It's nice that IBM has released OpenAFS two or so months after they said that it would be released, but a Free (libre) clone .
called ARLA has existed for sometime, and in my experience hasn't caused me any problems on several platforms, and is GPL'd
Also, arla supports many platforms, including (Free|Net|Open)BSD, and non x86 Linuxen, which Transarc (the IBM owned
company which actually develops AFS) hasn't bothered porting AFS to.
Arun
You can read about it here.
I found this interesting in their FAQ
"Will IBM support "Open AFS"?
IBM will support "IBM AFS" clients and servers for those customers who have active IBM AFS support contracts. IBM will not offer support services for Open AFS."
So... They're going to release it, and not support it. Joy...
IBM - "We're releasing this, isn't this great?!"
YOU- "wow, this is awesome, so how does bla bla bla..."
IBM - "RTFM!"
This sounds very much like a good thing; arla works okay as a client, but not great, and it's probably the least stable thing I have running under 2.4 right now.
I do have a few questions, though:
1) Does IBM own Transarc? What's the deal here?
2) What are the extra restrictions on the "IPL"? (Like we need YAOSSL (-> another license...))
---
pb Reply or e-mail; don't vaguely moderate.
pb Reply or e-mail; don't vaguely moderate.
I thought that Win2K had it's own journalling file system, so most users will just use that, irrespective of how good the alternatives are. IBM probably don't want to expend resources porting to a platform where it won't be used much, although there's nothing stopping some other clever folk from doing it.
Great, looks promising. Just a shame there's no win2000 support (well there's a surprise). I don't particularly wanna play with it on my linux machine, and I'm damned if I'm gonna resort to using NT4 ;)
...and I'll keep saying it: IBM is Linux's best friend, period. They're doing more to push Linux into the enterprise, and eventually on to the desktop, than any other company. This AFS release is just one more example of their commitment to Linux.
Someone moderate this bloke's post down as stupid.
Buddy, AFS might be old but that doesn't mean it is "dead".
AFS is a superior distributed file system which has a _proven_ track record. It has extensible ACL's. It has redundancy. It has fault tolerance. It is scalable. It has backing up built into its architecture. Kerberos fits nicely into the picture.
Let's say you have a large corporation, maybe you merged with other corporations. So now there's one corporation with all these departments that trust/might not trust eachother. Unix file permissions _break down horribly_ here whereas AFS shines. Just make groups for each corporation, add group names you trust to the ACL list of your directory and you're done. But that's not all, you can add individual users to your directory.
AFS is perfect for today's dot-coms who are now merging and forming huge corporations. And now that it is open source, it will be improved upon hopefully - not too familiar with the license.
Please read about AFS before posting ignorant-bad-big-corporation posts such as yours.
We should stand up and demand that they fully support Open Source by releasing code to viable products. If hundreds of thousands of programmers can do it every day we should expect the big guys to find a way to make it work.
lmao. Thank you for amusing me.
As someone has just pointed out on the Mailing List, Atheos' journaling filesystem is called AFS too. Now people are wondering what to call it to avoid confusion....
On my 2.2.16 box:
/usr/src/linux-2.2.5-15 does not exist.
/usr/src/linux-2.2.10 does not exist.
/usr/src/linux-2.2.12/include/linux/version.h does not exist.
/usr/src/linux-2.2.12-20 does not exist.
/usr/src/linux-2.2.13 does not exist.
/usr/src/linux-2.2.14/include/linux/version.h.
ERROR: Cannot build for Linux kernel 2.2.5-15:
ERROR: Cannot build for Linux kernel 2.2.10:
ERROR: Cannot build for Linux kernel 2.2.12:
ERROR: Cannot build for Linux kernel 2.2.12-20:
ERROR: Cannot build for Linux kernel 2.2.13:
ERROR: Cannot build for Linux kernel 2.2.14:
No UTS_RELEASE string found in
ERROR: Should be able to build at least one of 2.2.5-15 2.2.10 2.2.12 2.2.12-20 2.2.13 2.2.14.
Valid headers not present for any Linux kernel.
________________________________________
Napster-to-go says "Fill and refill your compatible MP3 player", which is a lie. It's not MP3. It's WMA with DRM.
So, can anybody give me the skinny on any Free (libre) DMAPI/HSM work going on?
I don't get it. First they scrap their plans for the Crusoe laptop, then the open source AFS. And then they use an obscure prprietry license on their open sourcing which makes it next to useless for most Open Source apps.
Do they support Open Source or not?
Yes, yes ,yes, thank you
demand? it's their software. buy a controlling share of IBM, then you can demand stuff from them. although i do see the appeal in your logic. i hereby demand that you hand over your bank account, credit cards, and any associated PINs. hundreds of thousands of bank customers have money they don't "need," they'll find a way to make it work.
Since AFS began at CMU prior to going to Transarc, it seems relevant to compare it to what came next. Coda is now the distributed filesystem pet project at CMU. Can someone compare/contrast the two?
Coda also appears to be at least partly GPL (or LGPL?), since it shows up in the kernel configurator. Was this a reaction to AFS going for-profit?
The living have better things to do than to continue hating the dead.
That's cuz there is no "IBM Open Source" license. OpenAFS is released under the IBM Public License.
Also, their lawyers probably had to go through all the code and clear it. That's why it took so long for us (SGI) to open source XFS.
Go Badgers! -- #include "std/disclaimer.h"
... Beowulf clustering?
Thank you.
That is a strange thing to say.
I feel a strong need for something like AFS when there is a environment with lots of different hosts, servers and clients.
It is as far as I know the greatest distributed filesystem with superb features like the backupsystem (which is like snaphots you can mount in your filetree)
Just the thing that you can easily have a common filestructure for the whole company wherever you are and not dependent on what OS you are running...
I do not share your view and you probably should tell us if you have had experience with AFS.
Probably not.
Why just Red Hat 6.2?
What about Red Hat 7, SuSe, Mandrake and, elementary, DEBIAN???
The nice thing about Windows is: it does not just crash; it displays a nice little dialog box and let's you press 'OK'
IBM isn't doing this? What about:
- SashXB (LGPL)
- JFS (GPL)
- Jikes (IBM Public License)
Or how about all the money IBM is pouring into Linux?- IBM invests $200M in Asia-Pacific Linux
- IBM Linux Wristwatch
- IBM's Overview of Linux (for investors)
- IBM banks on Linux
This has been but a selection of articles I could find in five minutes.---
How am I supposed to fit a pithy, relevant quote into 120 characters?
sorry, my bad. It is free, but incompatible with the GPL. http://www.gnu.org/philosophy/li cen se-list.html
Is it? I don't see it on that list. I see the IBM Public License listed as GPL incompatible, but that doesn't mean that the IBM Open Source license is the same.
Cheers,
Toby Haynes
Anything I post is strictly my own thoughts and doesn't necessarily have anything to do with the opinions of IBM.
The IBM Public License ("IBM Open Source Licence" is a typo/lack of understanding) is fully endorsed by the OSI. It was originally called the Jikes Compiler License, but when corporate wanted to use it for more projects they renamed it. You'll notice that Jikes is included in many of the current distros, hell, it's even in main debian and if that doesn't fit your definition of free I don't know what will!
No need for profanity.
Osi does not endorse free (as in liberated) software. They endorse OPEN SOURCE software. The GPL is a free software license, and as such is not compatible with the IBM public licence 1.0
---
Desperation is a stinky cologne
I've never understood this attitude. Especially in the context of the article, this strikes me as extremely ungrateful, rude and even childish. Something about Gift Horses and Mouths springs to mind.
You seem to be saying "Large companies whose business models include the concepts of selling and servicing software should immediately release their entire source code to the world at large". Without getting into the ethics, or the value of one business model over another, this attitude appears to be saying that the whole world should just stop what it's doing and obey the commands of a particular group of people.
Open Source / Free Software is a wonderfull, valuable, empowering movement. It's not the totality of the field, and it probably never will be. When corporations whose entire mindset involves the concept of exchange of cash for goods or services rendered embrace even a fraction of the values of these movements, it is indeed a cause for celebration. Not a time for beating them over the head that they haven't come all the way over from the Dark Side.
--
I'd rather have a bottle in front of me than a frontal lobotomy
Because of the licencing, this can't be included in the core distribution of our favorite Linux flavours.
It's open, but not free.
---
Desperation is a stinky cologne
And we shall have ezmail once again!!!!!!!
aahhhhhhhhhhhhhhh!!!!!!!
heavenly ezmail how I have missed you
Can't you see past the "release" of the product forwards to the benefits that this can have.
AFS may well be a dead file system but there is no doubt that in it's open source state it will prove an invaluable teaching tool - not only for people that are interested in seeing how a file system works but also for people interested in seeing how IBM have written the code.
Yes it is a good bit of publicity for Big Blue but let's not let disdain for the motives of IBM's marketing department blind us to the other oportunities that their actions have presented.
Matt Thompson - Actuality - Insert product here.