Anthony Towns has suggested I should use Blosxom for my blog; various people have asked for RSS feeds and permalinks, so it seems like a reasonable idea. So I looked at it, and there are several problems. The first is that it wants to run on the server, which is a complete waste of cycles: blogs should be generated statically and uploaded. You can run it in that mode, but then it generates broken URLs (I can force it, but all URLs should be relative so I can move my site and it doesn't break) and, mysteriously, gets the time on the entries wrong (no, it's not GMT, either).
Finally, it's aimed at "random thoughts" kind of blogs, not an online diary where the most important thing is the date on which the entry was written. So one entry per file doesn't really make any sense for this page, and a title line doesn't really make sense either. I don't really want to rewrite it, or write my own, so I'm debating what to do. One option is to simply run a script over this diary to produce the Blosxom entries, and generate a "pretty" diary from that. Of course, I like the idea of categories, so I'd have to insert markers in my blog for that.
Redid hotplug CPU fix against Linus' tree. More module-init-tools work. Random boring crap day.
Alli has turned into a big geek! Beautiful still, sure, but we were up until 1am ripping her CDs to FLACs and writing the scripts to convert them to OGGs, doing blind tests on what quality setting to use, and generally doing minor geek things. So slept in this morning.
Kim Weatherall mentioned me in her (excellent) blog re: Tradgedy of the RProxy . I feel famous!
Alli was poorly after returning from her trip, so spent a quiet weekend. But while she was away, I did get a lot of work done. Mind you, I ate KFC one night, and McDonalds the other two nights. Super size me!
Spent the evening ripping our CD collection, in violation of Australian copyright law (we don't have fair use, so I am a pirate), in preparation for Alli's Rio Karma.
As a glutton for punishment, I listened to the oral arguments of the (US) Ninth Circuit MGM vs Grokster case, which ruled that Grokster was not liable for introducing a new technology which can be used to infringe copyright (ie. the Sony Betamax case being retried). Of particular note is this exchange (25 minutes 10 seconds in) between the copyright holders' lawyer and the Judge:
Lawyer: the Internet is not a license to steal. There's nothing different from what they're doing from organizing and instructing the participants in a trafficing network that is trading in counterfeit materials, than in what they are doing.
Judge Noonan: Can I just... let me say what I think your problem is. You can use these harsh terms, but you are dealing with something new. And the question is, "does the statutory monopoly that Congress has given you reach out to that something new?". And that's a very debatable question; you don't solve it by calling it "theft", you have to show why this court should extend a statutory monopoly to cover the new thing. That's your problem. So, address that, if you would. Rather than use abusive language.
Indeed, the text of the article as summarized on copyfight reflects the judges clear understanding that the plaintiffs are doing everything but offering legitimate alternatives.
Ross Gittens did an excellent article on some of the problems with software patents, which reminded me to publish my article on The Tragedy of the RProxy.
Lots of kernel work cropped up in the last couple of days. I finally read my trivial patch monkey mail, and did a run at that. I also read through sched.c and suggested a minor cleanup, reworked the release_task path to fix a new bug introduced in hotplug CPU in the -mm tree, which hopefully is much cleaner now, and redid my old "DRM shouldn't use the deprecated inter_module functions" patch.
With my gorgeous wife away on a work trip, I am incredibly disorganised. I have remembered to feed the cats and replace their litter, but all other maintenance has been put on hold while I work a stretch of 16 hour days and generally enjoy myself. Despite its occasional frustrations, I have the best job in the world.
Meeting with Kate Lundy went OK; I think she knows where we stand.
Am meeting with Kate Lundy on Friday; the Labor party have made noises about copyright issues, and I wanted to make sure she has some idea of the "Linux" viewpoint. This is probably slightly different than "IT Industry" viewpoint, as we don't have to worry about losing Sony etc. as customers.
I started an internal IBM blog. Those of you inside the Big Blue Wall can read it here . My plan is to keep both it, and this diary, more regularly. Really.
Matthew Rimmer indicated that the Democrats are talking about fair use; I've expressed previously that simple "home copying" is defacto legal anyway, but some wider definition of fair use could actually bring about real benefits (such as legalizing PVRs like the Tivo).
I'm going to be giving the keynote at Linux Kongress in Erlangen next month, and attending the netfilter summit in the days before. Currently planning a "straight in, straight back" trip.
I referred my mother to my diary; I really do think that it's a useful resource to keep up with people that you don't see day-to-day.
So we lost the FTA battle. My wife is relieved it's over, at least, and she's decided she prefers to live in Canada than New Zealand. The question now is, what can we do to mitigate it? My thoughts, influenced by Anthony Towns, include:
The right to make copies for time and space shifting is the most common exception raised, but we already do that, legal or no, so I don't really care.
More importantly, I believe, the right to send a new copy between two parties who both already own a copy. This allows new models of distribution uncontrolled by the copyright holder, which we badly need.
Treaty or no, we have to find a way of allowing circumvention for access: otherwise control of technologies falls into the hands of copyright holders. Without this there goes the (legal) Linux desktop.
The treaty extends copyright another 20 years, when the wide consensus is that we should be shortening it, leaving even more works to lie fallow: their owner forgotten or uncaring, but noone else able to use them for fear of a lawsuit.
The solution I'd like here would be a copyright registration after 20 years: works not being actively sold and not registered, would have a cap of $1 damages for all "infringements". If the copyright holder wakes up, they can register their copyright, supply notice in writing to those actively distributing without a license, and those distributors have 6 months to stop or negotiate a license. This "phase out" period is required to provide business certainty: if you dig out an old forgotten song, and popularize it, the copyright holder shouldn't be able to return and shut you out.
163 pages of Australian legislation for the Free Trade Agreement; and that doesn't cover the details of ISP liability and anti-circumvention, which are still to come. Changes to 11 Acts other than the Customs Act.
The US Act HR4759: 32 pages. And theirs reads like a Trade Agreement: it's all about Customs and Tariffs, not a single change to domestic law.
I met with the very metrically-named Kelvin Watt who is the Labor candidate for my local seat (Eden-Monaro), to discuss the Free Trade Agreement IP issues. Too little, too late, probably. But he seemed nice enough.
Among other things, I gave him a draft of my expanded anti-circumvention litmus tests paper: the implementation of 17.4.7 of the FTA is going to be the next battle.
Back home, finally. On my TODO list is to read the proceedings of the Ottawa Linux Symposium , since I didn't get to all the talks I wanted.
Flew in to Portland yesterday, for Linux Networking Summit. Too bad Jamal isn't here, but I've got to say that 8 people works pretty well.
First day was pretty awesome; effective, even. Stephen Hemminger will be putting the summary up somewhere soon.
Interesting piece on the mathematical/Computer Science view of copyright entitled What Colour are your bits? by Matthew Skala, via Martin Pool's blog.
Thinking of implementing laws to ban circumventing Technological Protection Measures in your country? OK, here's three important benchmarks which any legislation should measure against to avoid suppressing competition, particularly from Open Source solutions. These issues have become clear since the Digital Millennium Copyright Act (1998 USA) and Digital Agenda (2000 Australia) were enacted.
"The Taken", Martin and Keith's band, played their first gig at the Pot Belly in Belconnen. It was actually good.
Good meeting with DCITA yesterday; these people can't do anything about the text of the FTA, but if we fail to highlight the dangers sufficiently and the FTA goes through, they will be drafting the anti-circumvention provisions (17.4.7) legislation. Unlike the rest of the FTA, we have two years to implement 17.4.7, but this is where our largest direct problems lie.
There are two litmus tests, in my opinion, for these provisions. The first is, if there are two devices which do the same thing, such as two different pieces of DVD playing software, and first is "authorized" and the second isn't, is it illegal to use the second one, and is it a criminal act to create it or distribute it? This is the Open Source DVD player test, which the US DMCA fails, and a straight-forward reading of the FTA text also fails. Our current Copyright Act also fails the second part of this test, but with the Digital Agenda Review recommendation 17, we would pass. This is a requirement for competitive software in this part of the industry.
The second test is: if I access a protected work in a way unintended by the author (such as skipping over those "no-skip" zones in DVDs), it it illegal, and is it criminal to create or distribute something which allows you to do so? This is a more fundamental question, one our Copyright Law probably passes: since a technological protection measure must "prevent or inhibit the infringement of copyright", and no-skip zones do not. You could imagine a less-clear case where a technological protection measure protects copyright and inherently places other access restrictions, but we pass the clearest case. Digital Agenda Review recommendation 17 would pass all such cases, since you have to actually infringe copyright to break the law.
The FTA does not pass this test, because legal protection is granted to anything which "controls access... or or protects any copyright". The United States law would be similar, but I am not a lawyer, and don't know whether a court in the US would consider their broad "fair use" precendents to override here.
Martin Schwenke complained that the snd_cs46xx module doesn't come back from suspend: this is required for the Thinkpad we use. The current work around is to unload and reload it across suspend, but it should work. But, despite spending a few reboots on it, it still doesn't.
Martin Pool made an excellent point recently, about rproxy and software patents:
I think rproxy is a particularly good example of just how much work there to get from a concept to a widely useful implementation. ("99% perspiration") The idea is clever, but you can think it up in a day. Implementation takes far longer: working out how to cope with all the nits in http; working out how to make it fit into Mozilla, Apache, etc; writing and testing the code; debugging, and so on.
Patents let people do the first one day's work and then charge rent on all the rest.
Alli updated her blog; I figure if we both remind each other we can actually be more regular about this.
Since we're planning the Big Finland Trip for March next year, and I'm going to be seeing Harald before then anyway in Portland and again at OLS, I decided not to go to the Netfilter Summit and Linux Kongress. However, Jos Vos just emailed me and said they'd intended to ask me to speak, and so asked if I would go anyway. It's hard to justify unless there's some actual customer work in Europe around that time.
Friday's session at IPRIA was interesting, but nothing hugely novel. But discussion afterwards (with, among others Peter Eckersley got me thinking hard about the requirement for raising awareness over the whole "irrational expansion of IP" issue. The search for a label for people who want to restore the public interest, however, reached "commonists" before I decided that naming things is hard.
I didn't read mail again until Saturday night. Big mistake, as someone discovered I broke module-init-tools on read-only filesystems. Damn POSIX locking: damn you to hell. 3.1-pre3 is running through tests now.
So I'm going to IPRIA's FTA IP session on Friday, to meet the negotiators and get some more feel for how these things work. I've also spoken with Martin Quinn of (Dept. of Foreign Affairs) about our concerns; their policy seems to be to push hard questions off as "implementation questions". So, I've got a meeting with Simon Cordina of Department of Communication, IT and the Arts next week.
New module init tools release: 3.1-pre2. Decided to try fcntl locking on modules: the old modutils did this to avoid races between loading and unloading. I dislike it because it's not foolproof (loading a different file with the same module name, for example). The original kernel code had no problems: one simple semaphore. Unfortunately, some modules try to load other modules during initialization, causing a lockup, so we have to drop the lock.
Anyway, Marc Boucher reported spurious errors when his module is loaded (it looks like two USB devices, so there are two hotplug scripts invoked which both try to load the same module). I tried just inserting a call to open and lock the module file, and in testing, it didn't work. Spent about an hour with strace, gdb, debug code... on the way out, I asked Stephen Rothwell if he had any ideas. He asked if the same process opened and closed the file elsewhere, as that would reset the lock. Damn good call; that was it, and POSIX file locking is really really braindead. An unrelated open and close in the same process will cause all locks on the file to be dropped.
Stewart Smith helped put together a page containing the testimony I gave it to the Senate committee; it serves as a good introduction to the problems with the AU-US Trade Agreement.
Alli and I are planning a trip to Finland. In winter (February/March).
See, she wants to see the Northern Lights. Establishing that they were only visible in very cold places in the middle of winter put her off for a year. But now the idea is back, and it looks like it's easiest to get it over and done with. Hence Finland, which is supposed to have better weather than Sweden for seeing the lights. Then our plan is to head south to Italy to thaw out.
Two pieces of FTA news:
I'm not entirely convinced it's that bleak. I think we can enact something better than the DMCA. Not better in the sense that it has more exceptions (that we can't do, and that I am complaining about vehemently - and that's the 17.4.7(e)(viii)/(f) problem), but better in the sense that the exceptions we do have are less technical, less narrow. The DMCA has a lot of additional language that the treaty doesn't have, and I think that means there are some things we can (and should) fix. We can define certain things more broadly.To some extent, I have to defer to this wisdom. For example, if 17.4.7(e)(i) is interpreted broadly (and clearly) enough in legislation, we can prevent the "Microsoft sues developer for reverse engineering Word 2007 files" nightmare:
Your other points, I entirely agree with.
(i) non-infringing reverse engineering activities with regard to a lawfully obtained copy of a computer program, carried out in good faith with respect to particular elements of that computer program that have not been readily available to the person engaged in such activity, for the sole purpose of achieving interoperability of an independently created computer program with other programsBut pushing the boundaries leaves two problems: firstly, we risk a trade dispute for not obeying the "spirit" of the treaty (the US clearly intends DeCSS to be banned as it is in their country). Secondly, we get to find out where the real boundaries are by litigation, which plays against Open Source developers, who tend to be small.
I've said it before; I don't fill in this page often enough. When I started, I could pick an event and find out when it happened by searching my diary. Now, where's the link to the time Chris decided that (in true engineer fashion), it couldn't be that hard to cut someone's hair, and cut Tridge's? Or even the recent TV news story on OzLabs, featuring various politicians speaking with my rubber chicken hanging clearly in the background?
I promised Alli I'd spend the week after Monday's appearance before the Senate Committee on the FTA, not talking about the FTA. I ploughed into the Trivial Patch Monkey and dumped a whole lot of patches on Andrew Morton, who forwarded them to Linus last night without a pause.
Even better, I got to spend some time with Alli today, just hanging out. She's cool, and I love her. And she should update her diary more often!
Last night I met with a politician who will remain nameless. However, anyone who can use the word "fistfucking" in polite conversation is someone to be admired.
I'm heading off to Singapore on Sunday for about a week, and will be spending a day in KL as well. I've only ever visited Singapore airport, so I'm looking forward to it.
Three points about the FTA 17.4.7 (anti-circumvention) I have come to realize:
Thanks to Kim for educating me on the first and third points.
I sat through the first half of the IP roundtable that the Senate committee organised this morning. Lots of lawyers, lots of subtleties.
By contrast, my own appearance concentrated on the issues for us as Open Source programmers and retailers: criminalization of DeCSS creation or distribution, illegality of Open Source access to things you have bought, and the chilling effect on small business of the lack of clarity (vs. normal copyright law: if you're copying, see copyright act, otherwise, don't).
I like my friends. They came over Saturday to help me prepare for tomorrow's appearance, including handling specific questions. Tridge provided pirate diagrams, and Martin Pool stayed up until midnight with me working on the (third!) iteration of the diagrams to use. This is a complex subject, and explaining it to someone not in the industry is difficult; the assistance of all my friends and colleagues is heartwarming.
I spent this afternoon with Matthew Rimmer who has done such impressive (and clear) work on Copyright (his JSCT submission on copyright extension is a must read IMHO), and with another IP lawyer, Kim Weatherall who has been tracking Chapter 17 of the FTA closely. She has legal familiarity with the DMCA which was really useful; she pointed out some things I didn't know, which deserve a page to themselves.
Politics is boring.
Sitting in a public hearing for the Senate Committee on the AU - US Free Trade Agreement, full of men and women in dark grey to black suits, light coloured shirts and boring ties for the men; two exceptions are myself and one other person.
Someone kill me.
The government previously commissioned a study (by the Canberra-based Center for International Economics) on the potential benefits of an AU-US FTA: answer, $US2bn ($AU4bn at the time, now only $AU 2.8bn). I can't find a copy of this report any more. They commissioned another study once it the treaty was completed, and the answer came out at $AU6.1bn, despite the exception of sugar and most dairy (which reportedly contributed some $800m to the original estimate). This is a little odd, to say the least. The CIE's statement on Lateline that "there's a 95 per cent chance, in our analysis, that the gains are between $1.1 billion per annum and something like $7-and-a-bit billion per annum" sounds quite confident though, doesn't it?
Looking through the report reveals two things. Firstly, the vast majority of these gains are not from increased trade, but from "investment liberalisation": they particularly concentrate on the foreign investment reporting we currently require. Increasing the notification limit from $50m to $800m is part of the FTA, yes, but if it's such a good idea, clearly we should do it anyway.
Naturally, I waded into the report to find what cost they had marked on the provisions that threaten Open Source development: the antircircumvention clauses. "These changes effectively reinforce the existing rights of copyright owners". So we're clearly not dealing with IP lawyers here, since these laws add new rights (this is the core of the debate: traditional copyright doesn't say anything about how I use your work if not copying it). Their conclusion: "It is not possibly to quantify the extent of any such cost". (page 54).
Let's cast the net wider on Chapter 17 issues. Extension of copyright? "Once again it is difficulty to quantify the extent of this effect as it is not possible to identify what proportion of existing material would be utilised to produce new works if the copyright term was not extended". Also: "However, it is not possible to derive any indication of the magnitude of the costs that may stem from the restriction of new works being produced from existing works".
These acknowledged but "unquantifiable" costs were ignored in the analysis (in fact, the report could find no sectors adversely effected by the FTA). Given the magnitude of the industries these 'copyright' laws effect, and the fact that US companies are by-and-large the incumbents here (ie. change is almost certainly in Australia's favour), this error margin dwarfs the gains anticipated by the report. Dampening our progress towards Open Source alternatives to (US) software companies alone could do it.
I'd love to see an economist's opinion on this report (ok, that's a low blow: we know the worth of economists' opinions in general).
FTA senate committee submissions closed today. I sent mine off, basically endorsing the Digital Agenda Review and Recommendations which restricts the anti-circumvention clauses to only cover copyright, and contrasting it with the direction the AUSFTA text, which criminalizes any circumvention of an "access control device".
So the Linux Australia FTA page got slashdotted yesterday, which was great. But the petition still only has 2100 signatures. Given that the european petition against software patents got over 300,000, I would have expected at least 10,000 signatures opposing locking in of universal patentability and DMCA-like access control laws.
I think there are three factors here. One, awareness is low, and Australians just don't care as much (this worries me: signing the petition isn't exactly a big committment). Two, our petition was less focussed. And three, there's been a lot of pro-FTA propoganda, like the talk about the $4 billion a year that the FTA was predicted to benefit Australia. Except that report was the top-end estimate, based on an idealized FTA where we got all agriculture tarrifs removed, etc. The independent report commissioned by the government (which kind of makes its independence a little suspect to begin with IMHO) has been delayed. Too late for the Joint Standing Committee on Treaties anyway: submissions for the FTA closed last week.
Work-wise, I've been cleaning up arch/ppc64/kernel/prom.c. It's a real mess, for several reasons. Firstly, dealing with Open Firmware is always non-trivial, and doing it from a kernel which isn't in the expected location yet, and without any real memory allocation routines, is horrible. Secondly, the code has grown incrementally, and it's the middle of a stable series so it's not a good time for a complete rewrite.
qemu-fast in 0.5.3 doesn't work: I have to kill -9 it. This is delaying the Debian release.
There's a big part of me which is grateful that the deadline on the Senate FTA committee submissions is the end of the month: politics makes me feel like I need delousing, and I hate confrontation. I get almost as upset about having been forced into confrontation than the anti-competitive clauses themselves.
This is placed here simply to assist David Gibson's bogotridge score.
The Linux Australia "How To Help!" page is up. In particular, there's an online petition which needs (tens of?) thousands of signatures if anyone federally is going take us seriously.
Linux Australia draft FTA position paper is out.
We have been far too quiet, and now we're in trouble. Help!
Awareness seems to be increasing that the "Free" Trade Agreement is not Open Source friendly . It's fairly clear to those in the IT industry, that as a nation of IT consumers, our interests are in more competition, rather than adding more powers to copyright.
Meanwhile, the work goes on. Hotplug CPU code is in, and we're playing a little with the toy i386 version. But mainly I'm looking hard at memory hotplug for PPC64.
Long discussion with Anthony Towns about the US-Australia Trade Agreement: much of it about the horror of existing laws. It's been an education for me, at least.
Discussion with an IP lawyer revealed that DeCSS is already illegal in Australia, as is region-freeing someone's DVD for them. It's just unenforced (the latter for fairly clear economic motives, since retailers have more difficulty selling region-locked DVD players). Normally I'm apathetic about unenforced laws, but the US DMCA experience has been sobering.
The main point here is that this treaty makes it much harder to fix our laws in the future: something which will probably come to a head as Free/Open Source software continues to make inroads and someone gets told they can't play DVDs they own on the laptop DVD player they own.
Several people asked me to update: I've been slack.
The recent Trade Agreement text came out yesterday, and there have been several discussions about it. It's a depressing document for Free Software, hence I don't use the term "Free" when referring to it.
We know who's on the senate committee: interestingly Ron Boswell , Nationals Senate Leader, is on it. He's from Queensland, so there might be an opportunity for small OSS businesses to talk to him about the concerns. It'd be nice to coordinate this though.
A week ago, Dan Shearer was in Canberra and teed up meetings with Kate Lundy (Labor Shadow Minister for the Arts, Sport and Information Technology) and an advisor to Aden Ridgeway (Democrats) who is organising the Democrat's response to the treaty.
We had excellent discussions with both over our fears on the FTA. The good news was they were interested: noone wants to be seen as opposing Linux & Open Source. What scared me most is that we had to explain things we pretty much take for granted: US patent system flawed, monopolies on simple ideas harmful, anti-circumvention "copyright" laws allow companies to impose extra (porfitable) restrictions by tying to a copyright measure, etc.
Martin Schlemmer (gentoo) reported a module-init-tools bug yesterday. By the time I got around to it, I found I couldn't reproduce it. Finally went to bugs.gentoo.org looking for more information, and posted a plea for the exact .config. But noone replied before I went to bed.
This morning, Jonathan Heaney replied, and I still couldn't reproduce it. I guessed that /proc reading was wrong somehow, and so installed a lot of random other modules: voila, reproduced problem, even though my guess was completely out.
I'd changed my "insmod()" routine inside modprobe to stop recursing if a module was already installed: it seemed an arbitrary but logical choice: surely if a module is installed, all its dependencies must be installed already? But I missed on vital point: if A depends on B and C, don't stop at B just because it's installed. If B depends on C too, that works, but otherwise C might not be installed. In some configurations, this actually happens.
Love valgrind, but it slows down my tests, and hence my release times have increased (it takes a full 20 minutes to run the testsuite on my laptop). Oh well, -pre8 out now.
Australia Day holiday yesterday; spent a quiet day. Ported hotplug CPU code to the -mm tree, which raised questions about integrating with Nick's new scheduler code.
Some more debugging of my qemu tetrinet server: qemu doesn't like a closed fd as input. /dev/null works better. Also upgraded qemu since a bug in 0.5.1 stopped the kernel from booting. Fabrice got straight back to me to report that (1) it was a qemu bug, and (2) it was fixed in CVS, which has since been released. Awesome.
Lots of time spent at linux.conf.au . The early part of the week was spent helping on the registration desk (I even got a gold helper's T-shirt on Tuesday) and ducking into various minicons.
The Linux and Open Source in Government miniconf was well attended, and a split from the rest of the conference, which is a developer con. However, it got some press coverage and increased attention. I also see the Open Source legislation as thawing the ground, rather than an end in themselves.
I didn't speak this year, but I did get to hold the auction for the T-Shirt, which was fun. I've got to say, as several other people did, that linux.conf.au is the most fun of all the Linux conferences.
The day of the dinner, I asked some speakers to send me any humorous emails they had to scroll on the screen during dinner. Unfortunately, there weren't too many given how late I left it, but I filled with various ones I've collected and shown at different talks. People asked that I put them up somewhere, as the projector was a little small for a room of 450 dining people.
I also received a hardcopy of LOTR from Linux Australia, for my role as Grandfather of linux.conf.au. Appreciated by completely unneccessary.
Machtelt Garrels from the Linux Documentation Project points out that their copy of the IPchains HOWTO now links to a porn site. That's the second time that's happened to an ipchains-related URL.
Finally got back to my trivial folder. There's a fair amount of stuff there, and even more spam.
Sent the new kthread stuff to Andrew Morton. Partially to give it wider testing, and partially because the hotplug CPU code requires it, and I want to get those patches into 2.6 fairly soon. But the other reason is that the code does make things much simpler, and I expect that once they are in, people will use them.
...on the other hand, sometimes ongoing debate means neither one is correct. I rewrote kthread again: now it's much neater, as well as being less invasive and easier to migrate to.
I also sent Andrew Morton a cleanup patch for workqueue.c, where it tried to catch SIGCHLD but never actually did, and in fact doesn't need to.
kthread implementation arguments. Sometimes someone simply disagrees with the way you've done something. And they produce a patch, and spend lots of time arguing. And you still think they're wrong. It's a real problem. If I'm argjuing against Linus, I'd just cave. Against Ingo or other people I know are better coders than me, I'd probably bow in favour of their arguments even if I disagreed with them, after some grumbling.
As it is, I've spent an inordinate amount of time walking through each kthread implementation decision, while I could have been coding something else. Eventually, it's used up it's timeslice and there's no choice but to regard it as closed, and move on. But I like consensus too much I think.
Kernel stuff has started moving again: lots of patches went into 2.6.1-rc1 from the -mm tree. Andrew took my
I wrote a script to grep lkml for patches adding interesting (ie. module-related) or obsolete constructs: caught a few pieced of new code using list_for_each() followed by list_entry. I obviously haven't publicized list_for_each_entry() enough, although it's now in 2.4 as well.