New Linux Development Model
Hi, posting from the outside so, no nice quotes and stuff.
Anyways, I was also miffed that the kernel folks merged a 'ancient' version of ipw2200 and ieee802.11, if they had merged something more current everything would have worked out of the box and all the cleanups would have been easier to cope with. Ie, the intel ppl could release straight patches to the in kernel version. I dunno if they have changed the way their driver works now.
Atm, the 'ancient' ieee802.11 is what breaks the ipw2200 build. So, basically all testing of cutting edge kernels gets very tedious due to the ieee802.11 package removing the offending .h file and making reversing -gitX and applying -gitY a real PITA.
Also, I dunno if it says what version of the firmware it should have...
Anyways, current state of ipw2200 is sad, and ieee802.11 is more developed out of tree. So imho at least ieee802.11 should be merged more closely and then generalized if needed (i think i saw some patches like that) and preferably ipw2200 should be merged as well.
This b0rkage has actually kept me from using wlan AND testing kernel.org kernels.
PS. Above, wrt merging, i assume that the API isn't quite 'stable' and that changing any drivers using them should be simple and straightforward.
stable as in, used all over and require major rewrites to change =) DS.
PPS. Why merge a outdated version of a driver/protocol/thingy at all? DS.
Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
Quoted 11 lines · Ian Kumlien
Anyways, I was also miffed that the kernel folks merged a 'ancient' version of ipw2200 and ieee802.11, if they had merged something more current everything would have worked out of the box and all the cleanups would have been easier to cope with. Ie, the intel ppl could release straight patches to the in kernel version. I dunno if they have changed the way their driver works now.
Atm, the 'ancient' ieee802.11 is what breaks the ipw2200 build. So, basically all testing of cutting edge kernels gets very tedious due to the ieee802.11 package removing the offending .h file and making reversing -gitX and applying -gitY a real PITA.
Those are no "ancient" versions, they are the "stable" versions of ieee80211, ipw2100 and ipw2200. ipw* folks think, and I have to agree, that for the stable kernel (Linux tree) it makes sense to add the stable versions of their projects.
For more about their versioning, make sure you read
http://ipw2100.sourceforge.net/#downloads .
Signature · 2 lines
Marcos Marado <[email protected]> Novis ISP
Quoted 14 lines · Marcos Marado
Quoted 11 lines · Ian Kumlien
Anyways, I was also miffed that the kernel folks merged a 'ancient' version of ipw2200 and ieee802.11, if they had merged something more current everything would have worked out of the box and all the cleanups would have been easier to cope with. Ie, the intel ppl could release straight patches to the in kernel version. I dunno if they have changed the way their driver works now.
Atm, the 'ancient' ieee802.11 is what breaks the ipw2200 build. So, basically all testing of cutting edge kernels gets very tedious due to the ieee802.11 package removing the offending .h file and making reversing -gitX and applying -gitY a real PITA.
Quoted 4 lines
Those are no "ancient" versions, they are the "stable" versions of ieee80211, ipw2100 and ipw2200. ipw* folks think, and I have to agree, that for the stable kernel (Linux tree) it makes sense to add the stable versions of their projects.
Yes, that would make sense if everyone interested changed the 'unstable' version instead, since they change the version merged it only creates problems. And if you DO check out the ipw2200 site you see that there has been no stable release for about a year and that there has been a large amount of bugfixes.
Also, the 'stable' version didn't use a separated ieee802.11 stack.
So to summarize:
Merging the latest version of both and then, if someone has problems, tell them to downgrade would be simpler. This also means that the ipw2200 team could release patches against the kernel as well as standalone modules.
The 'stable' version that got merged is more or less useless to people who are smart about their wlans. And on a side note, even the firmware has improved since then.
Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
Quoted 3 lines · Ian Kumlien
The 'stable' version that got merged is more or less useless to people who are smart about their wlans. And on a side note, even the firmware has improved since then.
People who is smart about their wlans is smart enought to update their kernels to use the new and possibily more unstable version. What's the problem? People maintaining the main tree may have a different (and valid) view on this issue. People is allowed to have different opinions than you, deal with it.
Quoted 26 lines · Ian Kumlien
Quoted 10 lines · Marcos Marado
Quoted 4 lines · Ian Kumlien
Atm, the 'ancient' ieee802.11 is what breaks the ipw2200 build. So, basically all testing of cutting edge kernels gets very tedious due to the ieee802.11 package removing the offending .h file and making reversing -gitX and applying -gitY a real PITA.
Those are no "ancient" versions, they are the "stable" versions of ieee80211, ipw2100 and ipw2200. ipw* folks think, and I have to agree, that for the stable kernel (Linux tree) it makes sense to add the stable versions of their projects.
Yes, that would make sense if everyone interested changed the 'unstable' version instead, since they change the version merged it only creates problems. And if you DO check out the ipw2200 site you see that there has been no stable release for about a year and that there has been a large amount of bugfixes.
Also, the 'stable' version didn't use a separated ieee802.11 stack.
So to summarize:
Merging the latest version of both and then, if someone has problems, tell them to downgrade would be simpler. This also means that the ipw2200 team could release patches against the kernel as well as standalone modules.
Quoted 3 lines
The 'stable' version that got merged is more or less useless to people who are smart about their wlans. And on a side note, even the firmware has improved since then.
I totally disagree. See: for those who don't crawl on lkml, don't compile kernels or modules or stuff like that, they had two choices: be without ipw2100 or ipw2200 or learn how to put the drivers in their kernels. Now, with the stock kernel you have ipw* support, even if limited for some uses. Most people will be happy with this version, but yes, there's still work to be done. When there's a new version considered stable it will get merged into the kernel. Until then, if you want to ride the unstable horse, you'll have to patch it yourself into the kernel.
If you want to simplify the process of building the unstable versios of ipw* or if you think that the newer versions of ipw* should be considered the new stable, or if you at some point disagree with ipw* development model you should complain in ipw2100 mailing list at
http://lists.sourceforge.net/lists/listinfo/ipw2100-devel . Kernel-related, the decision of supporting the latest stable is good and justifiable.
Signature · 2 lines
Marcos Marado <[email protected]> Novis ISP
Quoted 11 lines · On Wed, 2005-11-09 at 13:03 +0100, Wed, 9 Nov 2005 13:03:07 +0100 wrote:
Quoted 3 lines · Ian Kumlien
The 'stable' version that got merged is more or less useless to people who are smart about their wlans. And on a side note, even the firmware has improved since then.
People who is smart about their wlans is smart enought to update their kernels to use the new and possibily more unstable version. What's the problem? People maintaining the main tree may have a different (and valid) view on this issue.
Since it out of tree, it's not a issue of 'updating' a kernel, it's more of a issue of making it compile with a kernel.
(which if you read the original post was one of the problems mentioned.)
And, yes, they might have a valid view, but the kernel also has "experimental" features, why not merge a driver that can be used, mark it as experimental until it's merged with a stable rel. The issue of getting the first rel in to the kernel is that all of the sudden more developers starts hacking away on it and the original developers have to keep it all in sync (as i mentioned before).
Quoted 1 line
People is allowed to have different opinions than you, deal with it.
And that applies to me to, now deal with it.
Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
Quoted 3 lines · Ian Kumlien
Quoted 1 line
People is allowed to have different opinions than you, deal with it.
And that applies to me to, now deal with it.
That's easy - you are welcome to maintain whatever tree you like according to your opinion. Same as everybody else.
Repost, forgot the CC etc lines. (Yeah i know, reply to all etc)
Quoted 11 lines · Marcos Marado
Quoted 9 lines · Ian Kumlien
So to summarize:
Merging the latest version of both and then, if someone has problems, tell them to downgrade would be simpler. This also means that the ipw2200 team could release patches against the kernel as well as standalone modules.The 'stable' version that got merged is more or less useless to people who are smart about their wlans. And on a side note, even the firmware has improved since then.
Quoted 9 lines
I totally disagree. See: for those who don't crawl on lkml, don't compile kernels or modules or stuff like that, they had two choices: be without ipw2100 or ipw2200 or learn how to put the drivers in their kernels. Now, with the stock kernel you have ipw* support, even if limited for some uses. Most people will be happy with this version, but yes, there's still work to be done. When there's a new version considered stable it will get merged into the kernel. Until then, if you want to ride the unstable horse, you'll have to patch it yourself into the kernel.
Since it has to be out of tree the way it is now, you don't patch or upgrade, you simply have to make it compile. And if you are happy with wlan driver that hardly does wep then you have some issues as well.
Quoted 7 lines
If you want to simplify the process of building the unstable versios of ipw* or if you think that the newer versions of ipw* should be considered the new stable, or if you at some point disagree with ipw* development model you should complain in ipw2100 mailing list at
http://lists.sourceforge.net/lists/listinfo/ipw2100-devel . Kernel-related, the decision of supporting the latest stable is good and justifiable.
Thats why we have "experimental" drivers in the kernel?
like, f.ex.:
Packet Engines Yellowfin Gigabit-NIC support (EXPERIMENTAL) New SysKonnect GigaEthernet support (EXPERIMENTAL)
Or the protocols? (STCP, DCCP, etc)
IF a unstable version of a driver is better for the user isn't it better to merge it as experimental instead of merging a old version that wreaks havock on users of newer kernels (and they will STILL run the unstable version on older kernels anyways).
Note: I use skge myself, so pointing out that they are experimental is in no way negative, i actually prefer it to the stable driver.
Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
Quoted 8 lines · Al Viro
Quoted 3 lines · Ian Kumlien
Quoted 1 line
People is allowed to have different opinions than you, deal with it.
And that applies to me to, now deal with it.
That's easy - you are welcome to maintain whatever tree you like according to your opinion. Same as everybody else.
ROTFLOL, Yeah i know, I just think it's more about the users need than 'stable' and 'unstable', Just label it experimental so the user knows what they are getting into. (Just like the examples posted in other mails)
Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
Quoted 16 lines · Marcos Marado
Quoted 11 lines · Ian Kumlien
Anyways, I was also miffed that the kernel folks merged a 'ancient' version of ipw2200 and ieee802.11, if they had merged something more current everything would have worked out of the box and all the cleanups would have been easier to cope with. Ie, the intel ppl could release straight patches to the in kernel version. I dunno if they have changed the way their driver works now.
Atm, the 'ancient' ieee802.11 is what breaks the ipw2200 build. So, basically all testing of cutting edge kernels gets very tedious due to the ieee802.11 package removing the offending .h file and making reversing -gitX and applying -gitY a real PITA.
Quoted 4 lines
Those are no "ancient" versions, they are the "stable" versions of ieee80211, ipw2100 and ipw2200. ipw* folks think, and I have to agree, that for the stable kernel (Linux tree) it makes sense to add the stable versions of their projects.
To what end? The current drivers compile and load, but they don't function for wireless communication! What's the point of having code which is essentially a no-op, why have it if it doesn't provide any functionality?
With the current firmware and driver a "scan" shows 14 connectible points outside an apartment building (only one secured in any way ;-) whic is just what Windows shows. With the stock kernel zero are found. That's not stable that's moribund.
Quoted 4 lines
For more about their versioning, make sure you read
http://ipw2100.sourceforge.net/#downloads .
Signature · 3 lines
-bill davidsen ([email protected]) "The secret to procrastination is to put things off until the last possible moment - but no longer" -me
Quoted 4 lines · Bill Davidsen
With the current firmware and driver a "scan" shows 14 connectible points outside an apartment building (only one secured in any way ;-) whic is just what Windows shows. With the stock kernel zero are found. That's not stable that's moribund.
Sorry to disagree, but I use the stock kernel version of ipw2100 almost daily, with no problems.
Signature · 2 lines
Marcos Marado <[email protected]> Novis ISP
On Wednesday 09 November 2005 16:01, Ian Kumlien wrote:
[snip]
Quoted 10 lines
Quoted 9 lines
I totally disagree. See: for those who don't crawl on lkml, don't compile kernels or modules or stuff like that, they had two choices: be without ipw2100 or ipw2200 or learn how to put the drivers in their kernels. Now, with the stock kernel you have ipw* support, even if limited for some uses. Most people will be happy with this version, but yes, there's still work to be done. When there's a new version considered stable it will get merged into the kernel. Until then, if you want to ride the unstable horse, you'll have to patch it yourself into the kernel.
Quoted 3 lines
Since it has to be out of tree the way it is now, you don't patch or upgrade, you simply have to make it compile. And if you are happy with wlan driver that hardly does wep then you have some issues as well.
Eh? WEP seems to work here with different key sizes. I think most people should be satisfied with finally being able to use their wireless controller in ANY form on a Centrino laptop without having to find a driver!
Quoted 14 lines
Quoted 7 lines
If you want to simplify the process of building the unstable versios of ipw* or if you think that the newer versions of ipw* should be considered the new stable, or if you at some point disagree with ipw* development model you should complain in ipw2100 mailing list at
http://lists.sourceforge.net/lists/listinfo/ipw2100-devel . Kernel-related, the decision of supporting the latest stable is good and justifiable.Thats why we have "experimental" drivers in the kernel?
like, f.ex.:
Packet Engines Yellowfin Gigabit-NIC support (EXPERIMENTAL) New SysKonnect GigaEthernet support (EXPERIMENTAL)
Quoted 6 lines
Or the protocols? (STCP, DCCP, etc)
IF a unstable version of a driver is better for the user isn't it better to merge it as experimental instead of merging a old version that wreaks havock on users of newer kernels (and they will STILL run the unstable version on older kernels anyways).
I'm not sure what you mean by "wreaks havoc". There have been bugfixes, but they do not affect many general use-cases. Equally the WPA work really IS brand new.
The primary "havoc" I had when upgrading to 2.6.14 was finding the older firmware and co-installing it.
Quoted 2 lines
Note: I use skge myself, so pointing out that they are experimental is in no way negative, i actually prefer it to the stable driver.
I really don't think this is your call. The driver maintainer can push whatever code he's most happy with for whatever reasons he likes. I'm confident that in time appropriate changes from 1.0.8 will make it into mainline.
Perhaps the only criticism here that is valid is that Intel haven't released a 1.0.9 that builds against release quality (2.6.14) kernels _yet_. But at best that is a topic for another mailing list, not the LKML.
Signature · 6 lines
Cheers, Alistair. 'No sense being pessimistic, it probably wouldn't work anyway.' Third year Computer Science undergraduate. 1F2 55 South Clerk Street, Edinburgh, UK.
Quoted 8 lines · Marcos Marado
Quoted 4 lines · Bill Davidsen
With the current firmware and driver a "scan" shows 14 connectible points outside an apartment building (only one secured in any way ;-) whic is just what Windows shows. With the stock kernel zero are found. That's not stable that's moribund.
Sorry to disagree, but I use the stock kernel version of ipw2100 almost daily, with no problems.
I concur, ipw2200 is also fully functional here. This sounds like a genuine bug, or a mis-configured system, not a criticism of the current Linux development model.
Signature · 6 lines
Cheers, Alistair. 'No sense being pessimistic, it probably wouldn't work anyway.' Third year Computer Science undergraduate. 1F2 55 South Clerk Street, Edinburgh, UK.
Quoted 2 lines · Alistair John Strachan
Eh? WEP seems to work here with different key sizes. I think most people should be satisfied with finally being able to use their wireless controller in ANY form on a Centrino laptop without having to find a driver!
Err.. there are lots of different APs out there, and the in-kernel borken driver doesn't work for me with several APs I need to access.
Not to mention that WEP doesn't provide any meaningful security, whereas the out-of-tree driver even does WPA2 with ease.
The big problem for users, is that they used to be able to just go and grab the *real* ipw2200 driver, and build/install it with a few simple commands.
But because the kernel now contains all of this way-outdated stuff for ieee80211 and ipw2200, there are tons of header file conflicts and protocol mismatches when attempting to build/use the real driver.
Sure, us kernel folk can cope with all of that (in theory, though in practice I'm still stuck with 2.6.13 because I haven't yet gotten working ipw2200 with 2.6.14, with *either* driver).
But things just got WAY more complicated for most users of ipw2200. Sure, they can ignore us and just continue to run their old vendor kernels. But this means they don't get up-to-date kernels with bug fixes and security fixes. And more importantly to LKML, we've now just cut off a potentially large crowd of kernel-testers.
Ugh. Ugly.
On Thursday 10 November 2005 15:10, Mark Lord wrote:
[snip]
Quoted 3 lines
Sure, us kernel folk can cope with all of that (in theory, though in practice I'm still stuck with 2.6.13 because I haven't yet gotten working ipw2200 with 2.6.14, with *either* driver).
I'm probably asking a silly question here, but presumably you've grabbed ieee80211 1.1.6 and ipw2200 1.0.8 from their respective websites, compiled them, installed them (whilst running 2.6.14) and the resulting driver does not work?
Quoted 7 lines
But things just got WAY more complicated for most users of ipw2200. Sure, they can ignore us and just continue to run their old vendor kernels. But this means they don't get up-to-date kernels with bug fixes and security fixes. And more importantly to LKML, we've now just cut off a potentially large crowd of kernel-testers.
Ugh. Ugly.
I completely agree with this assessment, I was merely defending the "linux development process" which I do not believe to be at fault here.
Signature · 6 lines
Cheers, Alistair. 'No sense being pessimistic, it probably wouldn't work anyway.' Third year Computer Science undergraduate. 1F2 55 South Clerk Street, Edinburgh, UK.
Quoted 8 lines · Alistair John Strachan
On Thursday 10 November 2005 15:10, Mark Lord wrote:
[snip]Quoted 6 lines
But things just got WAY more complicated for most users of ipw2200. Sure, they can ignore us and just continue to run their old vendor kernels. But this means they don't get up-to-date kernels with bug fixes and security fixes. And more importantly to LKML, we've now just cut off a potentially large crowd of kernel-testers.
Quoted 4 lines
Quoted 1 line
Ugh. Ugly.
I completely agree with this assessment, I was merely defending the "linux development process" which I do not believe to be at fault here.
It isn't, this started as a mail from someone, and was continued because ipw* and ieee* things was mentioned. I really hope that a newer version is merged... It can be branded experimental and all... =)
Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
Quoted 7 lines · Alistair John Strachan
Quoted 6 lines · Marcos Marado
Quoted 4 lines · Bill Davidsen
With the current firmware and driver a "scan" shows 14 connectible points outside an apartment building (only one secured in any way ;-) whic is just what Windows shows. With the stock kernel zero are found. That's not stable that's moribund.
Quoted 6 lines
Quoted 2 lines
Sorry to disagree, but I use the stock kernel version of ipw2100 almost daily, with no problems.
I concur, ipw2200 is also fully functional here. This sounds like a genuine bug, or a mis-configured system, not a criticism of the current Linux development model.
Given that using recent firmware and drive software works and stock doesn't, I personally have a hard time saying that the configuration is wrong. Being old-fashioned I conclude that the thing you change to fix a problem is most likely the thing that caused the problem in the first place.
can't speak to the 2100, I have a 2200, and with the current firmware and driver it works.
Signature · 4 lines
bill davidsen <[email protected]> CTO TMR Associates, Inc While an unsigned integer can never be negative, it can be less than zero for sufficiently large values of zero.
Quoted 15 lines · Bill Davidsen
Quoted 7 lines · Alistair John Strachan
Quoted 6 lines · Marcos Marado
Quoted 4 lines · Bill Davidsen
With the current firmware and driver a "scan" shows 14 connectible points outside an apartment building (only one secured in any way ;-) whic is just what Windows shows. With the stock kernel zero are found. That's not stable that's moribund.
Quoted 6 lines
Quoted 2 lines
Sorry to disagree, but I use the stock kernel version of ipw2100 almost daily, with no problems.
I concur, ipw2200 is also fully functional here. This sounds like a genuine bug, or a mis-configured system, not a criticism of the current Linux development model.
Quoted 8 lines
Given that using recent firmware and drive software works and stock doesn't, I personally have a hard time saying that the configuration is wrong. Being old-fashioned I conclude that the thing you change to fix a problem is most likely the thing that caused the problem in the first place.
can't speak to the 2100, I have a 2200, and with the current firmware and driver it works.
You need to use *exactly the right version* of the firmware (2.2 I believe) with the ipw2200 driver in 2.6.14. Non-2.2 firmwares just don't work with the 1.0.0 driver (though the latest firmware obviously works with 1.0.8).
Are you using WPA{,2}? The driver in 2.6.14 doesn't do WPA, only WEP.
Signature · 6 lines
Cheers, Alistair. 'No sense being pessimistic, it probably wouldn't work anyway.' Third year Computer Science undergraduate. 1F2 55 South Clerk Street, Edinburgh, UK.
Quoted 4 lines · Marcos Marado
On Wed, 2005-11-09 at 14:05 -0500, Bill Davidsen wrote:
Quoted 9 lines
Quoted 3 lines
With the current firmware and driver a "scan" shows 14 connectible points outside an apartment building (only one secured in any way ;-) whic is just what Windows shows. With the stock kernel zero are found. That's not stable that's moribund.
Sorry to disagree, but I use the stock kernel version of ipw2100 almost daily, with no problems.
Well if ipw2100 works on one machine for one person, obviously ipw2200 must be correct as well. I never said they didn't work anywhere, and I have no old 2100 hardware.
But the new {firm,soft}ware works and gives the same results as XP, while the default says there are no usable signals in range. And like it or not, it has to work at least as well as Windows on the same machine to be considered functional. Many devices work better.
Signature · 3 lines
bill davidsen <[email protected]> CTO TMR Associates, Inc Doing interesting things with small computers since 1979