As I don't have connectivity in the new flat yet, I'm looking for a sane way to read my email via GPRS. I can live with server deleted emails still being visible if it can save me having to transfer one megabyte of headers on opening a folder (like mutt does). Since I pay by transferred volume and not by time, it would fetch mails lazily. Anyone know of something that works?

Well, that’s it, folks. People who helped to make it happen are too numerous to mention here, but I would like to specially thank Joshua, Marc and Andres for guidance and mentoring, another Marc for being a great AM, and Joerg and James for doing awesome DAM work.

The transition of my main desktop machine to udev seems to be complete. Thanks to the people that offered useful suggestions to my previous post. The only additional problem I had was the loop module not being auto-loaded, causing Xen to fail. (I use loopback files for the Xen instances. Less efficient, but easier to manage, etc). Friday was my last day of working at Noha-Systems / Intasys / Acecomm Solutions, so I’ve got the weekend off and then I start working at Bytemark on Tuesday. (Monday is a bank holiday in England & Scotland) Happily this means I can stay out late on Sunday, dancing naked on a hill at The Beltane Fire Festival on Sunday Evening. Result!

To all those who emailed, called, /queried me: Thanks. Your support is really appreciated.

To all those who shared similar experiences with me: thanks. It s good to feel understood, to feel that I am not the only one that gets torn apart for the loss of a small furry friend.

There was a huge blackout in Madrid yesterday. No subway, nice traffic jams, fights for taxis (I witnessed) and, of course, servers going down. Carlos took care of the remains of Emacs and yours faithfully in the midst of a wonderful Downtime: Lilo not booting, all is bad issue in his biggest customer. H ctor and Bego run upstairs (4 floors, that s true love), and comforted Vi and me. A very special, tall, blond, skinny, wonderful g ero spent most of the afternoon on the phone with me. Even Chicho, my ex, called and cried with me. Thank you all.

Emacs had a happy life, she was a spoiled cat, a very happy and very spoiled cat. She lead a (short) good life, a much better life than, sadly, too many humans. I am starting to accept her loss and feel some peace, after a dreadful night.

I called the vet today to find out how to ease the pain for Vi, the whole team was devastated with the news. Their advise was to get another furry feline friend for Vi as soon as possible, and even told me about a two months old that s up for adoption, but it s not the best moment right now. I ll be leaving for Mexico in a week, and I am not comfortable with the idea of leaving two furry strangers alone for two weeks. I already feel terrible about leaving Vi for so long, specially now, but H ctor and Bego will, again, take good care of him and spend plenty of time at my apartment to keep him company.



There s laundry hanging still out, and I can t get to open the fucking window.

Amaya -- so sorry about Emacs. I had two cats for a couple years, and lost Felix during a move back East, The saddest thing has been that Leo lost a brother to play with and had to become sort of a loner, and rely more on human companionship. Last night Leo suprised me, I was idly swinging my headphones around and he began to play with them and we had a good old style tussle like we used to when he was a kitten, and I thought, hmm, I should probably get him some proper toys again, cats never really outgrow that even if they get to go outside and hunt small furry creatures and butterflies and annoy other cats. Wish his brother was still around.. Anyway, I think you should still take Vim to the beach sometime.

She was the hunter, the jumper, the smallest one. She was the first to join me in bed every night. She was the one that sat on my arms when I was at the computer, the reason of all my misspellings on IRC. She was the one that preferred peeing on the bathroom basin than on the litter box. The one completely crazy about getting in the bath after my shower, the one who liked to pour the water out of her drinking dish. The one that purred the loudest, the one that sucked on my clothes, the one that made strange bird noises.

Today I came home from work and bad, terrible, news were there for me. I would have liked to blog about how much certain rubito makes me smile, or the Extremadura Sessions, or any of the wonderful stuff that has happened to me lately.

I was in such a good mood when I arrived home today. Well, I admit I was a bit pissed off. While waiting for a green light to cross the street a stupid pigeon shit on my sweater. But that doesn t matter anymore, because cats die stupids deaths and Emacs had jumped after another stupid pigeon, from a seventh floor, through a window that was open only because the lady that cleans my house was doing the laundry I had been to lazy to do myself. Nobody s fault. But she s gone. And I loved her. She would have turned one year old in July, but cats die such stupid deaths. She had to jump out of that window and I am so angry I want to shoot every pigeon on the face of earth. I was one day late, I was building screens to protect the windows with, and this was the only screen left. I was going to finish that very window today.

Christopher Aillon says:

Well, I just got back from the Desktop Linux Summit to find an email awaiting me telling me that my proposal for a talk on NetworkManager was rejected this year for GUADEC. ... I can see why getting contributors into GNOME is rather difficult. Even though I'm going to continue working on GNOME because I'm paid to, it's rather disheartening to go through this. Looks like it will start to be a work-only thing instead of a work-and-free-time thing. To all you who have somehow made it into the "in crowd", have fun in Catalonia. I probably won't be going.

To be fair there were two Network Manager talks submitted, one by Christopher and another by Robert Love. What is worse: telling one of the two that their talk isn't accepted, or having two talks at GUADEC on the same subject? NP: August and Everything After, Counting Crows

In other words, we could now produce our 1948 standard of living (measured in terms of marketed goods and services) in less than half the time it took in that year. We actually could have chosen the four-hour day. Or a working year of six months. Or, every worker in the United Stares could now be taking every other year off from work-with pay.

-- The Overworked American In a way I've managed to make that choice. I currently work four hours a day on paid work; over the past 8 years that's been my average thanks to taking several years mostly off work. Of course things are really more complicated, since I spend a lot (often more than four hours) of my leisure time on technical "work" that's just as demanding (if less stressful generally) though not done for money, and since there's a definite bleed over between what I'm working on on my time off now and what I'm likely to be paid for doing in a couple of years time. The article suggests that we would need to cut our desired standard of living in half to work half time. While mine is probably lower than average, it's not that much lower, at least for this area; I have to make up the difference by being a more desirable employee than average, and part of how I do this is "banking" my contributions in my free time. This is an aspect of free software that many people don't seem to get, that your contributions can pile up over time and end up being personally beneficial later. Of course health care is really problimatic when working a reasonable amount of time, such as four hours, in America. You have to budget at least 3/4 your health care costs to be paid yourself, not as a benefit. Working in Europe begins to seem really attractive, etc.

The m68k port has been having issues with GCC 4.0 since its inclusion in the archive. Although many of the bugs that spawned from that update have since been fixed, there are still a lot of them; I suspect the high number of GCC 4.0 ICEs is the main reason the m68k port still hasn't made it back as a release candidate architecture. Many of the issues with GCC 4.0 seem fixed in GCC 4.1. The most recent example is QT4, which did not build with G++ 4.0, but did build with G++ 4.1. Another particularly problematic example is some KDE-related thing (the name of which I can't remember offhand) that didn't build with GCC 4.0 due to an ICE, and didn't build with GCC 3.4 either, due to a different ICE. However, it's hard to say for sure whether GCC4.1 really is an improvement if you haven't tried using it on an autobuilder. For that reason, I just installed gcc 4.1 in the buildd chroots on jazz and quickstep, the experimental autobuilders, and made them the default compilers. Hopefully that'll work better than GCC 4.0.

I feel a bit sad. Is FOSS really just going to be an alternative forever? Or can it break out of the trappings of alternatives, and make a world a better place? =) Instead of posting this to BiOS, I felt that it would be better posting it here; feel free to comment, feel free to ignore

Get a grip on reality. Even windows is an alternative to us FOSS loving peeps. Saying that FOSS is an alternative IMHO is a good marketing/advocacy tactic to get potential new users to get their feet wet on FOSS.

And now you are risking yourself to sounding hostile. At the risk of sounding like a bad guy, I have been shifting in between the planes of Ubuntu and Windows via cold water over the past few weeks; wala lang, I just wanted to be reminded of how it feels to be shackled’, so to speak. But really, I’m doing this because I can, and as a result I’m having a rocking time appreciating the subtleties of both Windows and Ubuntu So I guess I’m not so far removed from reality as you seem to portray. Now, let me play the simpleton, who’s really an astute participant. If a FOSS desktop, say Ubuntu’s Gnome, is supposed to be as easy as Windows, then why should I bother learning it at all, when I can simply transfer my old Windows skills to Ubuntu without much a problem? Another question: why should I be motivated to switch to FOSS at all? Perhaps you can dismiss this as FUD, but you are none the wiser when your reasons for switching to FOSS is only because that it is an alternative, that it is cheap (hard to get cheaper being free), and that its mandated by the local government. Among other things petty and cruel. Indeed, these are the questions people ask to us FOSS people, even long after the good shows (OSForum, SFD, LinuxWorld, etc.) have their run. As a matter of fact, I keep asking this to myself everyday, for the past 3 years of using GNU/Linux everyday. And every single day ends with an answer: “I use FOSS because it empowers me to surf,chat,email,program with relish.” For many other people, unfortunately, they don’t even get around to use a FOSS system, much more a CLI; thence they just make do with “I use Windows because that’s just what I have.” But I don’t want to force people to switch (or even just use) FOSS because of those reasons. I want them to use FOSS exactly because it empowers them, allowing them to get their jobs done effectively and efficiently. FOSS isn’t out there to deceive, as you so imply, maybe unwittingly. Well, perhaps, sometimes it is, but more often than not, its there so as to convey a deeper truth: that you can do things with it, while at the same time feel empowered, and FREE, with it.

Once they get their feet wet, it would likely be a natural progression for these users to realize the benefits that FOSS can offer (full-fledged, stand-alone and yet cooperative at the same time, approach to solving real problems). The key here is to just even entice the user to try, to take it for a spin, that would be a big step towards FOSS land.

Yes, getting feet wet is a good thing, it has its purpose, namely, for a person to acknowledge that he’s stepping on water, and preparing for a plane shift But from the point of view of a swimming instructor, getting feet wet is not good enough; it is better to get a person wet his head into the water, because that’s the time you know that he’s ready to learn how to swim, even if he doesn’t feel he’s ready. The same is true for newbies wanting to dive in like in that Linux Crash Course. It is not good enough that a newbie be exposed only to just one interface, just one true way (yeah, that coming from an Emacs guy I do not want them to slave away moving the mouse to and fro, doing some stuff under a GUI when they can just do the same thing in less time by just a few keystrokes. Perhaps, most wouldn’t be ready to use a CLI immediately to do some hand-hacking, but, like air in the planet, its always there, even when you don’t have an immediate need for it (air, however, is something we always have an immediate need for And neither do I recommend learning the intricacies of the shell, whether be it Bourne, Korn, or Peanut shell. Indeed, there’s nothing wrong about being at least bilingual; I want my audience to learn just a little bit of everything I want them to be aware, that there’s more to FOSS than what we’re just showing them. I don’t want them to deny the freedom and potential of looking what’s within a GNU/Linux system, unlike the usual situation of getting a brand-new PC with just Windows in it, with not a clue as to how things work within it. It’s like showing what’s under the hood of a car, but unlike the casual driver who only knows just how to drive it, the FOSS user is much more free to tweak whatever pleases; yeah, the danger of overheating is there, but with care, that could be addressed. In short: I don’t like this kind of false FOSS advocacy, when you’re effectively denying the chance for other people to improve and empower themselves, just so because CLIs seem to be so complicated to be taught in a single day No; I don’t think a shell has to be complicated and convoluted and inscrutable, like governments, among other things; rather, they are their own kind of fun, like finding a sleuth solving a case in a whodunit, or winning the World Cup. I’m no longer wanting to be in the business of long, impassioned, flamewars (is this thread one?) so I’m stopping here. Like I said earlier, it’s a speaker’s call to make adjustments on his presentations so that it reaches its target audience effectively and efficiently, just as it is a call for a FOSS user to use a GUI or the CLI to solve a problem. Besides, I have a swimming class to teach tomorrow Dive in!

Now that I've finished adding email subscriptions to ikiwiki the only thing blocking me finally making an official release of it is html sanitization. This is the part I've put off till last because it's the part I hate. Somehow the web has evolved a backwards security model that wants sites with dynamic content to be responsible for policing that content for things that could cause security issues, instead of just making browsers actually secure so that no possible html can be a security issue. Which is of course absurd, but the various types of potential cross site scripting attacks that users of your wiki will be vulnerable to if it doesn't try to sanitise its html are nothing to laugh at. So everyone is left having to sanitise any untrusted content that they put up on the web. In my experience, almost everyone does it poorly, I've even found XSS holes in well-designed websites like Advogato. It doesn't help that browser coders are busy adding new features^Wholes to their web browsers, which everyone then has to update their sanitisers to block. I don't like participating in security arms races, so while I am evaluating various html sanitisers right now, adding html sanitisation to ikiwiki is not a pleasant idea, and I am tempted to punt on it, document the issue, and refer anyone experiencing problems to the authors of their web browsers. Sigh.

Last year, I gave a talk on XML differencing at the Extreme Markup Languages 2005 conference in Montr al. Since then I didn't have much time to work on that project. The prototype app is opensource, and available at ssddiff.alioth.debian.org. The slides are there, too, and so is a link to the article. XML differencing is actually a two phase process; I've been mostly working on the first phase. Traditional approaches intertwine the two phases. In the first phase, you build a mapping between the two documents. I called this a "nodeset correspondence" in the slides. Basically you decide which node from the first document corresponds to which node in the second document. When you have found this mapping (the "marked" output writer of ssddiff will output this information in form of two XML documents marked with numbers in a certain attribute), the actual differencing is done. But much of the work remains. When doing a diff, the output (format) is just as important as the actual result. While the "marked" output explained above is very precise, it doesn't give the user an overview of what has changed in the document. That's why I added two more output writers, the "xupdate" writer and the "merged" writer. The "xupdate" writer does obviously produce an XUpdate script. It goes a long way to optimizing the output (read: minimizing the number of operations), and employs a longest-common-subsequence algorithm for that (LCS is what regular diff uses on lines!). Fortunately we already have a 1:1 mapping, so this boils down to a longest-increasing-subsequence which can be computed very efficiently. Then the writer employs some XUpdate tricks to do real moves (by storing parts of the tree in variables to be inserted later), and avoids collapsing of text nodes (it will eventually insert dummy nodes for that). But all this gives just a glimpse of the complexity I've encountered in what I considered to be a "remote" part of my project. It isn't even described in the paper, because I just hacked this code together to work somewhat (attribute handling is incomplete IIRC). The "merged" writer turned out to be much more powerful than I had imagined and actually rather easy. After "merging" two version of an HTML document, I could add some CSS rules and have a visual presentation of the changes in my browser. Very cool. I'd love to add this html merge functionality to the prototype, but I didn't have enough time yet (you'll find some refactoring of the merged writer in SVN though to get there someday). But let me get back to the difficulties in the "XUpdate" writer. Not all of these difficulties are actually of technical nature. It is (like with the semantics of differencing) a matter of the intended data model and use. When you are doing a diff to be applied by a computer (e.g. via XUpdate), you may have the requirement, that the resulting document is identical to the original document. In that case, the diff is likely to have to restore whitespace. If you are interested in analyzing the differences, you will want to skip any irrelevant information. Let's assume we have an address book:

<abook><entry name="tom" /><entry name="jerry" /></abook>

Ordering of the entry nodes is likely to be irrelevant. You address book applications will sort the entries on demand. So a "semantic diff" shouldn't include any changes to the ordering of "entry" nodes, should it? The XUpdate writer doesn't have any functionality in place to skip such changes from being written to the output file. This will likely be not possible without some DTD magic and extra user information. However, to have the benefits of the "-u" context option of good old "diffutils" - fuzzy matches, and somewhat reliable merging of changes without having all revisions available - such semantics are needed, too. But this will even further increase complexity of the output writer, which for example would need to detect somehow that the "name" attribute in my address book is a unique key, and therefore xpath expressions like /abook/entry[@name='tom'] are preferrable to expressions such as /abook/entry[1], which will easily break when merging. The specification of an "XUpdate-replacement with context information" is probably a whole thesis in itself, I guess. P.S. I'm occasionally asked about a Java version of ssddiff. Sorry that I have to disappoint you there. First of all, I'm very short of time. Then I'm not too comfortable in Java, and I estimated that the Java version would need at least twice as much memory, because of certain optimizations I can't easily access in Java that I can just hide in a template in C++. And memory is the key limiting factor at least in exact mode.

Please correct me if any of RSS/Atom/etc. do already have threading. On a mailinglist I'm on (large number of subscribers, and often threads become quite long but only interesting for like 1% of subscribers, but with days with now new mails inbetween) there is a discussion every now and then by people who'd prefer a forum. A forum however is a totally non-working solution. First of all, web UIs (even Ajax, Web 2.0) suck. None offers threading as comfortable and smooth as a real email reader such as Evolution or Thunderbird (Outlook seems to have quite bad threading, or disabled by default, thats why many novice users think forums do better threading...). Secondly, a forum requires you to frequently visit it, whereas email will patiently sit in your email folder waiting to be read, just like your regular email. That is much more workeable with an increasing number of lists/forums. You can easily be subscribed to 20 mailing lists. Can you imagine visiting 20 forums every day? Even with smart browser (e.g. Galeon that can open a bookmark folder into tabs, did I mention IE sucks?) this is quite a pain. Now imagine that some of these forums are inactive for weeks. Visit the forum just to see that there are no new posts? Now RSS comes into mind. But for a Forum? Should it list every post? No, it needs to have thread support, so I can easily kill whole threads. Or I'll just stick with email and tell everyone who wants a forum to fishing. Or write a forum with reasonable email support. Oh, and RSS doesn't solve the writing problem either. No, RSS is not the answer. Email is. Update: A couple of people have send me emails about NNTP/Usenet/News. While there are excellent clients for it available, and the distributed nature is also a nice feature, it also has drawbacks. For example, you basically need to setup a separate service (especially if it's a private newsfeed, and you can't use gmane), which is much more difficult that running a mailing list or a RSS feed. You can run a RSS feed with basically any provider. You can get a private mailing list with many. Then you usually need to setup a new profile for each NNTP server, because it's designed totally around the server model, not single groups. Furthermore, while you can usually access RSS feeds with your regular web browser (in form of their HTML counterparts, or some built-in filter), and mailing list with your email program, users need separate software for using News at all. (Granted, many email clients can do news) Oh, and then people seem to be afraid of NNTP, unfortunately.