Juri Pakaste - The Blog

2002-09-30T16:30:00.000Z

Moving on.

My weblog lives now at http://www.iki.fi/juri/blog/ (yes, I know it redirects to www.helsinki.fi, but www.iki.fi is the "official" location. For now. :-)

Thank you to blogger for a great service. I just outgrew it.

2002-09-26T14:34:00.000Z

Moving on?

I've been thinking about moving away from Blogger. Blogger makes it very easy to start a blog, but I want a bit more freedom, and nicer evironment to write in.

I can't really afford a web hosting solution at the moment, so that limits my choices considerably. I have an account the University of Helsinki, but they don't allow anything but static pages. For dynamic pages, Blosxom seems attractive. It does have it's shortcomings, but the basic concept sounds like something I want.

So what to do? I hacked up pyblosxom, the Python version of Blosxom, a bit, to support the features I need and to allow for easier static mirroring of the pages. The end result? My static blosxom blog!

I dunno if that will be the final solution, I'll have to play with it a bit. If it works out, I'll release the stuff and write a small howto.

2002-09-25T12:07:00.000Z

Brokenness in next/prev behaviour of Straw

I've been noticing that Straw's next/prev behaviour is a bit, um, erratic. I've blamed it on the fact that my local Straw DB is probably somewhat broken, due to me hacking on it all the time. It just reached new heights, and I decided it was time to investigate. Sure enough, a few minutes of digging uncovered some stupid mistakes, caused by changing some abstractions slightly and not moving everything that depended on them along.

This is now fixed, expect a 0.10.2 release this evening (UTC+3) or
tomorrow.

2002-09-25T09:59:00.000Z

Subscription placement

Brent Simmons thinks out loud about charging for default subscription places. If I were Brent, I'd go for it. No reason not to, as long as you retain your independence. I've been thinking about adding default subscriptions too; though I haven't come up with anything I'd put there besides Straw news and FootNotes (which would be a whole lot better if they put the complete articles in the feed :-)

I can't really do charging for the placement; Straw's popularity (and probably maturity, too) is far from NetNewsWire's, and doing that sort of thing in an open development environment where there will potentially be other people working on it too, people I don't really know, is rather dubious. I'd rather make a policy of not adding any extraneous feeds than wonder about every change if it was paid for by someone or not.

But NNW's development is more closed and it's a popular piece of software. It'd be no problem keeping track of them and there would be no questions of trust. I'd probably do something like mark them as sponsored feeds (think Google), keep a limited number of them at a time, and keep a clear policy about how long to keep them there and what you allow as content. And don't do anything silly like not allowing the user to remove the feeds. Keep things honest.

2002-09-24T10:15:00.000Z

Minor

Released Straw 0.10.1 last night.

Search is now a bit better. I didn't do hysteresis, but instead just made pressing keys in the find text field stop the rendering of the search results and start with the new results. At least with the systems I use, searching is quick enought, but rendering a list of few hundred items takes time. I also fixed some bugs in the search: it uses a pseudo-smart system of storing search arguments and results in a stack and checking on new search terms if they are a subset of some previous search, but the is_subset function wasn't working really very well in 0.10. It should be better now. Oh yeah, and search got a toolbar button too :-)

Mark Pilgrim has started to use content:encoded elements (RSS 1.0 land) in his channel items, so that the description contained only a short synopsis and the beef was in content:encoded. So now I support that. The current system is that content:encoded overrides description. I don't know if this is the "correct" way, I didn't really find out when reading the RSS 1.0 home page how they should be treated.

Straw also now reads the stuff in the blogChannel module of RSS 2.0. We don't do anything with it yet, though, I'll have to figure out what kind of UI I want around it first. And implement OPML reading and writing too, I guess.

The final thing is that the DB problems (which I hope I fixed in the 0.9 series) have occasionally caused empty items to be stored in the database. Straw doesn't break anymore when encountering those.

2002-09-20T10:41:00.000Z

Semantic Cyc

As I said, I don't worry all that much if we don't get the RDF stuff in RSS. I share some of the doubts expressed here.

When talking about metadata and making stuff "easy" for computers to "understand", people should always remember that this stuff is not easy, and computers aren't going to turn smart overnight, however much metadata is fed to them.

2002-09-20T08:09:00.000Z

New release of Straw out

As some of you might have noticed, there's a new release of Straw out, 0.10. It has two new minor features — number of unread shown on the title bar, button to sort feeds alphabetically — and one important-ish one, incremental search of items.

The title bar status view is bit of a compromise. What I really want to do is a status applet, or docklet, or whatever is the approved technology for this kind of thing in GNOME 2. But this does its job and is a nice solution, so it will most probably stay.

Search is a feature which will need some love before it's where I want it to be. There's at least one bug I know (add something in the middle of the search string, it won't show what it's supposed to), it doesn't deal with updating feeds and it's a bit too eager to start searching. As of version 0.10, it doesn't wait for one microsecond to start searching once you've inserted something into the search field. It should probably have a hysteresis of, um, maybe one third or one fifth of a second, so you could type normally and it would search when you are paused. Maybe it could deal a bit better with keyboard navigation too, and I'm still not quite sure should find pop up a dialog as it does now or should it just unhide the controls in some part of the main window.

But still, it's pretty nice. Even if it's not exactly like the search feature in Apple's mail application, which is what Marko wanted.

2002-09-18T10:33:00.000Z

blogChannel module

Thank you.

2002-09-16T13:05:00.000Z

On competition

Dave said Straw competes with his own software. I never really thought about it that way, but I can see his point of view. Two things about their relationship and the competition.

I've never used Radio, but it is a full-blown weblogging tool, for both authoring and aggregating content. Straw is strictly an aggregator. I have no idea what percentage of UserLand's Radio customers use it only for aggregation, and I don't expect them to tell, but still, the focus is different. I've been thinking about writing a desktop weblog authoring tool, with possibly some kind of communication going on between it and Straw, but that's a project that might never happen, and would be a separate tool.

The other thing is platforms. Straw is written for the GNOME desktop. Radio works on Mac and Windows. I don't have access to those platforms, and therefore Radio; users of those platforms don't (well, I suppose OS X users might theoretically run Straw) have access to Straw. GNOME does — or will some day — compete with those two desktops, and I see Straw as a part of that, and in a way, Radio is on the other side of that fence, as a part of the larger world of those two desktops. But that's desktop A versus desktop B, not software package Z running on A versus software package X running on B. But unlike me, Dave has a commercial interest in this, so I wouldn't be surprised if he saw things differently.

Of course, Radio might be ported some day, and so might Straw (no, that's not in the current plans and I have no interest in doing it.) If they do port Radio, I genuinely wish UserLand the best of luck with it, because I'd love to see them, and a commercial desktop software package for "normal" users running Linux, succeed. I don't know if the time is ripe yet.

2002-09-16T10:09:00.000Z

Thanks, everybody

It does wonders to your ego, being hyped by (at least) Aaron, Mark and Uzopia (which doesn't seem to answer at the moment, so no direct link.) And finally, Dave putting my face on Scripting News was quite a surprise :-)

(And just to clarify — or maybe not — my position in the whole RSS thing, I'm maybe not quite as pro-more stuff in base RSS as the bit in Scripting News makes me sound. I'd just like to see some new features in it, and I'm not enough of a Semantic Webber, and am sceptical enough about everything related to XML, to be too gung-ho about the namespaces thing. But namespaces might keep things more hygienic, so maybe they might be good. Shrug. Let's just get on with it.)

2002-09-15T21:04:00.000Z

Brokenness detected

There seems to be something strange with fetching new articles, figuring out which ones are new, and saving them with the current version of Straw. I haven't yet checked if this is problem is already there with 0.9.3 or with just my current development version, but I'm seeing definite weirdness here: Straw thinks that some oldish articles are new, and doesn't seem to save the articles quite correctly. I guess I should dig into this instead of hacking on search...

[UPDATE, +15 minutes]

Take a feed, read it through, add an item in the middle of the RSS, poll the feed, everything is suddenly unread. Whee.

[UPDATE, a few minutes more]

And I'm locking up the DB too, trying to quit straw results in a hang.

[UPDATE, too many hours later]

Phew, maybe now the thing works. 0.9.4 released.

2002-09-12T10:08:00.000Z

Solutions creating new needs

You fill one need, and the solution creates another. With Straw, I can effortlessly follow a far larger number of blogs than I could without it — at the moment, I'm subscribed to 34 feeds which are easily read through every day (or many times a day.)

But it's not easy to cope with the information load. Now I need more tools to help me with that. One thing I notice is that after I've read through all the blogs in the morning, I have a lot of trouble remembering what was where. I remember there was something interesting somewhere, but finding it can be a pain.

So next up, functionality-wise, will probably be search functionality (free text search, with time limits) and possibly something like bookmarks. Nice to know there's no shortage of stuff to work on.

2002-09-12T09:58:00.000Z

And another release

I don't think 0.9.2 really fixed the DB problem, 0.9.3 should be better.

2002-09-12T08:00:00.000Z

New release, new home

Last night I released version 0.9.2. No new features, really, just a bit of UI polish here and there, plus hopefully a fix for the DB problem I described yesterday.

Straw has also moved: it's now hosted at Savannah. Here is the project page. And here is the new home page.

The development of Straw will now be more in the open than before. This will also allow for easier participation by others. There's now an open bug tracking system, patch manager, etc, and I can give access to the CVS.

2002-09-11T19:23:00.000Z

Great literature

I usually don't much talk about books here, but one I just read is worth a mention: Neil Gaiman's American Gods, which apparently also won the Hugo for best novel this year, was an incredible piece of work.

Combine old gods and legends (you see here many of the same names which were seen in Sandman), new gods formed by the psyche of the modern man, and a touch of the american mythos, road movie style. The result is outstanding. I knew nothing about the book before I started reading, except the author; I never read back cover texts and I had heard nothing about this, but for the first time in quite a while, and from the very first pages, it was very difficult indeed to put this book down.

Highly recommended.

2002-09-11T19:05:00.000Z

Will code for food

Well, not quite. But my (ex-)employer Noitatieto finally, after months of struggling, went bankrupt. Which kind of sucks. Civil service takes still 8 hours per day, but a daily income of 12,5 euros doesn't allow for much of a lifestyle.

So, you. Yes, you. Employer in the Helsinki region. Need a part-time software developer? Buzzwords: Python, Java, Common Lisp, Perl, PHP, Javascript, C. Zope, J2EE. XML-RPC, SOAP, CORBA. Gtk, GNOME, Swing. Linux, Solaris. Relational databases. Five and a half years of work experience. Contact me.

2002-09-11T15:34:00.000Z

Eek

Seems the database of my local installation of Straw is pretty much FUBARed. I wonder what happened. Straw started to act a little weird after it got duplicate guids from Scripting News today; now Straw (or more properly, Berkeley DB) is telling me

DBRunRecoveryError: (-30989, 'DB_RUNRECOVERY: Fatal error, run database recovery')

and db3_recover says

db_recover: DBENV->open: No such file or directory

Not very much fun.

[UPDATE]

There was a definite bug in the handling of adding new items with duplicate guids, and I killed that one. I'm not quite sure though if this is the thing that caused the other problems, but it seems likely, even though I have no idea how.

2002-09-11T08:48:00.000Z

Perfectly simple

At artima.com, there's an interview/conversation with Ken Arnold (the JavaSpaces guy) by Bill Venners. The subject is "perfect design" (aka "The Right Thing To Do", preferably with the stupid capitalization) and simplicity in design.

There's some really nice insights there, about how a design's quality and usefulness is context-dependent (meaning: there's usually no one right design), and how when designing systems to use by programmers you should regard the programmers as users and the design as an UI problem. And also remember that not every programmer is equal: there are differences in skills, experience and the amount of attention they are able and willing to give to your system. Sometimes something complex can be good, and sometimes not.

The part about "can a design be too simple" reminds me of the Einstein quote: "Everything should be as simple as it is, but not simpler."

2002-09-08T17:18:00.000Z

Climbing to fame

Ha, http://www.google.com/search?q=straw gives the release announcement of Straw 0.4 from freshmeat as hit number seven, at the moment. The stuff that is before it is totally irrelevant, obviously.

2002-09-07T22:33:00.000Z

On UI

Toolbar. What is it for? For grouping often-used UI functions together, in an easily accessible place. Hmh, ok. Kind of like keyboard shortcuts for occasional users.

Now, what should it contain, in Straw's case? The current meager selection of buttons (preferences, poll) dates back to the time when there were pretty much just those two functions in the main window. Poll arguably belongs there. On the other hand, with background polling, a feature which has been there from day one and which I suspect is more used than manual polling by people who don't habitually break their aggregator and restart it thrice a minute, its usefulness is limited. Preferences most certainly is not an often used feature.

So what others? Um, the most often used feature I use is navigation between items and feeds. Usually by repeatedly hitting space, reading all the unread articles. But that's hardly a feature which warrants a toolbar button — if you don't know the keyboard shortcut, you can just navigate by mouse, in the normal GUIy point'n'grunt way.

Catchup? Maybe. Poll selected? Umm... no, I don't think so? Quit? No way. So, what else is left? The two new UI gadgets I added in 0.9 do need their buttons, in fact three of them, but those buttons are located elsewhere in the window, in places I felt were somewhat logical.

That pretty much exhausts the current functions. The usual suspects from document processing applications — new, open, save — don't apply here. We don't do documents or files, not in a way the user can see (which is why I'm not too keen on the File menu either, but sometimes it's better to be consistent with other applications than trying to be all bold and inventive.) Browserish back, forward, reload, stop don't really work for us either, even if they are a slightly better fit. Well, I suppose there could be use for "Stop" functionality. Search is one thing that will appear in the application and which could live there, but I'm not sure when will it happen.

So should I just junk the toolbar? The window looks a bit naked without it, doesn't it? Leave it as it is? I'm not too keen on that, either.

Oh well.

2002-09-07T22:01:00.000Z

.1

Oops, noticed just after release a display bug in preferences dialog, 0.9.1 fixes that. And I'm not going to announce this one on Freshmeat & www.gnome.org.

2002-09-07T20:58:00.000Z

Get it while it's hot

Fresh-baked Straw 0.9 is now available in the download directory, see the Straw home page for details.

Next up: a more detailed release announcement on the home page. And in the RSS feed. And on Freshmeat. And on www.gnome.org.

2002-09-07T20:11:00.000Z

Straw 0.9 soon

Look, some new snazzy UI: channel information display and a subscribe tool.

The subscribe tool uses Mark Pilgrim's rssfinder.py and is pretty smart.

There's been also pretty nice UI speed ups; I optimized quite a bit the feed list display routines (which were, basically, totally stupid) and also pruned out one extra signal emission when moving feeds up and down in the list.

This release will require version >= 1.99.13 of PyGTK and gnome-python.

2002-09-07T19:08:00.000Z

Motivations

Of course, the real reason I posted the last item, even though the wish is real enough, is to avoid monotony: I hope I'll be able to release Straw 0.9 later today, and it would have been a bit too much to have last Saturday's entries titled 'RSS 0.94' and 'Straw 0.8', and this Saturday's 'RSS 3.0' and 'Straw 0.9', now wouldn't it?

2002-09-07T19:01:00.000Z

More on Referer, and GNOME

One wishlist item for GNOME libraries: url_show() should accept user-defined headers as extra arguments. This would mean that more browsers would also have to accept them on the command line. I'm not sure if it would be very useful with any other headers than Referer, and I don't know do any other applications besides Straw need that, but still. It would be nice to have.

2002-09-07T10:41:00.000Z

RSS 3.0

Now we're really getting somewhere. I was, you know, all frustrated and stuff with the difficulty of parsing XML. This is what we really need.

2002-09-03T13:43:00.000Z

On Referer

John Wiseman complains about something I've noticed too: referer logs are getting a bit useless. The embarassing bit is that he lists Straw as one of the offenders :-)

I'd like to argue in my defense a bit. That particular URL isn't supposed to appear in Referer logs; it did while I was working on the feature, but now the one that goes there is similar (in fact the result of some severe copypasting) to Aggie's AggieReferrers.html: http://www.iki.fi/juri/straw-ref.html. And a pointer to the user's weblog, if it is known. If not, there's a bug in Straw 0.8 :-) There's three alternatives to this I can see.

Straw could just not create a Referer header, the way it was in 0.7 and before. But that would mean no information at all. The way it is now, you get to see who reads you, even if they don't comment on it.

Straw could put a direct link to the user's weblog, but then, that would mean more "false positives". It's more than likely that there would in fact be no link in the weblog, because the Referer header, identical to one generated by a web browser, would appear each and every time the reader refreshed their view of your feed.

Straw could generate the Referer header only when the user has, in fact, filled in information about their blog/personal web page. ATM, it's done always, only leaving out information about the user's weblog if it's not been supplied.

I think I'll go for the third option. John is right in that the User-agent header does the job of conveying information about the reader software quite nicely. But I think the information about who reads you is interesting, and by using the indirect link, you don't get the problem of false positives. By not generating the header without the link to user's blog, you don't get the extra noise.

2002-09-02T15:19:00.000Z

Always waiting for the next great thing

Nothing is ever quite as good as it should be. I'm always waiting for some piece of software to get released, usable, something.

Most of the things I wait for aren't all that important, or affect usability of things much at all; it's just about getting something new. Or making something old just a bit prettier or working a bit better in some respect, or being somehow more "powerful". Getting the new thing gives a moment of satisfaction, and using it is a pleasure, at least for a while. But then I'll be again waiting for the next enchancement, for the same gadget or for something else.

I guess this is what they call consumerism, even if I the software is free. Or maybe just severely lacking systems.

2002-08-31T15:03:00.000Z

Straw 0.8 is out

Various small fixes, Mark Pilgrim's rssparser.py plus RSS quality indication for the user, Timothy O'Malley's timeoutsocket. More information here, as before.

2002-08-31T13:35:00.000Z

RSS 0.94

Well, I have complained about this so many times in the past that I thought I should mention it: there's now a (draft) specification of RSS 0.94. It even documents all the elements from the previous RSS 0.9x incarnations.

2002-08-29T09:11:00.000Z

Nobody's perfect

Common Lisp is a wonderful language. It really is. If you don't know it, you can't judge it, and if you know it, you probably love it, if you have any taste whatsoever. I'd be using it all the time, if it weren't for, in some cases, other languages making it easier to deal with the rest of the world. But even the best of standardising committees do make their mistakes.

2002-08-28T14:36:00.000Z

Write

I usually don't much care for the meta-blogging discussions, but Jarno summed up at least some of my feelings about the why of all this writing. I just tried to explain yesterday to a friend why I prefer this to a medium such as usenet news, but I don't know if I really did it right (or if he was interested :-)

These things are fundamentally different. Blogging is about content production and writing. Usenet is about discussions. I really don't care all that much about long-winded threads going always over the same stuff; I want to write, occasionally commenting on something someone else has written, but mostly about things I have done or seen. I really don't want to worry about being on topic in a forum; anything I want to write is on topic here, by definition. I don't want to participate in the mindless flammage and meaningless meta-debates discussion threads tend to degenerate into. I want my writing to be always visible (as long as you know the URL), not hidden somewhere there among thousands of messages, expiring from a server message pool when too many other people have filled their HD space.

This is all about writing. And also about, egoistically, me, and the things I care about.

2002-08-26T13:58:00.000Z

RSS diversity

The multitude of RSS versions out there causes many problems. I have complained about that in the past. Winer apparently had an idea about merging the two main branches, RSS 0.9x and 1.0 (of course, this wouldn't help with the vaguely documented mess that is 0.9x.) He just doesn't bother explaining how having a flavor attribute would help if all the tags inside the containing element were still different, and how it would be better than just having different top-level elements which is the current situation.

2002-08-21T11:39:00.000Z

GNOME HIG: 130 pages of fun

The usability guys working on GNOME released human interface guidelines, version 1.0. It has been long in the making, in its various guises. Back in 1998, gnome-gui-list was the battleground of much useless fighting over something that had a name but otherwise never got far: The GNOME Style Guide. But things have progressed. Serious effort by people who know what they are doing has helped a lot here, and without having read the finished document, I believe this thing might be worth something. It's time for application developers to make their stuff comply, if possible, with the contents of that document — usability, accessibility and consistency are all good things, and should be easier to achieve with the help of the HIG.

And I guess the group "application developers" includes me.

2002-08-18T04:05:00.000Z

Nekkid

Center of Helsinki. 18th of August, 2002. 04:30 in the morning, sun just starting to rise. 1500 people, both sexes, all ages, totally naked. Spencer Tunick shooting pictures of you, for his Nude Adrift series. First picture of people standing up, second lying on the ground in the street, third sitting around the Havis Amanda fountain, fourth of just the men standing in the park in front of a statue. I was there.

It was one of the weirdest experiences of my life. It's not that nakedness itself is weird — most Finns are quite used to it, thanks to sauna — but the sheer mass of people, in a place where you are definitely not used to seeing people without clothes. You totally blend in the nude mass. It was fun.

[UPDATE]

So 1500 was the estimate Tunick gave when addressing us before we stripped. The reporters have been talking about 1900.

2002-08-17T09:10:00.000Z

Straw news available for your aggregating pleasure

Release announcements and stuff like that, in RSS: Straw news.

It's not quite identical to the news on Straw home page, because I'm updating the RSS feed by hand.

2002-08-17T08:13:00.000Z

Haaaappy biiirthday tooo youuu

Happy birthday to GNOME: courtesy of Footnotes, here's Miguel's original announcement five years ago.

It's been some trip, and my desktop looks and feels a lot nicer these days than in 1997.

2002-08-16T23:42:00.000Z

Release

Straw 0.7 is here. Stores items and images locally and stuff. Slightly better UI too.

2002-08-16T22:53:00.000Z

Gee, free ideas

It's nice when you don't have to figure out all by yourself. I've had some vague ideas about making addition of feeds in Straw easier (the subscribe to source button is first step in that direction, but few people use the source element) but haven't really thought about it all that much. There's a lot of functionality to work on, and something that works, even if not optimally, isn't as high on my to do list as something which is totally broken or unimplemented.

This is something Mark Pilgrim has been thinking about, though, even if not in the context of Straw. He is frustrated with current aggregators and his step-by-step instructions on how the process of finding a feed should work sound sensible enough. So maybe when I get around to working on this again, I'll take a clue or three from that rant.

Meanwhile, Straw's local storage of items and images (offline operation, here we come) seems to work. Maybe I'll get a new release out today/tomorrow.

2002-08-14T09:15:00.000Z

Learning things about Zope

So I was agonizing over getting SSL to Zope working. M2Crypto might be nice, but it requires SWIG, which brings with it a version-dependency nightmare. Supports SWIG 1.3.6, I have 1.3.13, it doesn't work. Oh, great. I'm not going to downgrade.

Next choice was in fact my first choice, before I started looking into more 'native' solutions: Apache SSL proxy. I've used it many times before with great success, so I figured it would be a piece of cake.

Well, it almost was. The only problem was, once I got it up & running, the first page loaded a-ok but all the links pointed to the supposedly-proxied Zope server instead of the SSL-enabled Apache proxy. I soon enough figured out there must be something funny what Zope is doing to cause this, but I had no idea what. The headers looked OK through telnet-ssl. Then I noticed in the HTML an element I had never heard of before: base, with a href attribute.

With kind help from people on #zope, I found VirtualHostMonster, a product packaged with Zope. This thing is really pretty cool: with a simple ProxyPass directive in Apache, I'm able to make all the objects inside a folder in Zope to think they live on a different host, behind a different protocol, under a different directory. And it fixes the base href too. And the VirtualHostMonster doesn't require any configuration at all.

So, ProxyPass / http://host:port/VirtualHostBase/https/vhost:443/path/to/stuff/VirtualHostRoot/
not only makes stuff on host:port appear on the Apache SSL vhost, it also makes things on the Zope located at host:port in the directory /path/to/stuff/ think they really are at vhost:443/ root.

Gotta remember this next time I'm reinventing the wheel.

2002-08-13T11:19:00.000Z

But I wanna be conservative

Jarno just nudged me in the direction of Mark Pilgrim's Ultra-liberal RSS parser. I've been a bit frustrated lately with everybody and their dog producing broken XML, so maybe I'll integrate that into Straw and see if it makes things any better. I must admit I'm not too keen on it, though: I'm familiar with the phrase "be conservative in what you send and liberal in what you accept", but still, I'd really love it if the leaders of the bloody interoperability movement would respect the standards they claim to use.

2002-08-12T09:04:00.000Z

They're at it again

Three new buzzwords for oh-so-useful product comparisons. "We can't use that, it doesn't have this, um, WS-Transaction thingy!"

2002-08-10T15:05:00.000Z

Kudos to Sleepycat

Reading the Berkeley DB Reference Guide is a refreshing experience: a manual for a product the company has built its business around, and they actually say in many places that the system is not ideal for every need. This is something you don't see every day: usually even the technical documentation for various products is so hype-filled — or at least totally silent on the downsides — that you'd think the thing was written by God himself.

2002-08-10T10:38:00.000Z

HTML tricks: forms

HTML is a wonderful thing. Well, no, it isn't, really. But still, sometimes it can be pretty nice, if you put in the effort and embrace new standards. Dan Loda's Doing forms justice does a nice job of showing how to do more accessible forms with CSS.

Two things about this, though: I'm not convinced shortcuts in forms are really relevant. How often do you find yourself filling the same forms many times, anyway, and not doing in linearly? Shortcuts become useful when you learn them, and that requires practice. But I suppose there might be situations where that could be more than just markupbation.

The other thing is that all that markup looks rather painful to type without some serious help from the editor. I'm no big fan of PSGML mode in Emacs (even if it does mostly behave after some fighting) and I'm not convinced I'd enjoy using it to do that form.

2002-08-10T01:58:00.000Z

Things you notice

When creating and using a RSS aggregation tool, you notice some things.

You are able to follow a whole lot more blogs than before.

Because of that, you start to look for more interesting blogs.

And notice there aren't too many of those, after all.

Could be I'm just lousy at using Google, too.

Blogs without RSS feeds, like this one, start to annoy you.

Blog/diary hosters without automatic RSS feeds, such as Blogger and Advogato, start to really annoy you.

Free software is nice, Raph Levien of Advogato told he would be happy to integrate a patch to mod_virgule which would allow RSS feeds from diaries. I'll be looking into this.

Movable type with its default of putting only n first words of each post in the RSS feeds starts to annoy too. I'd like to read the complete posts, thanks.

Places such as Footnotes and No Logo which only supply titles in RSS feeds get on your nerves too. I wonder if this is good for my blood pressure.

People don't really respect XML standards all that much. Neither does certain software, such as Radio, which is apparently happy to include ISO 8859-1 and Windows characters in nominally UTF-8 encoded feeds, as well as HTML entities. Interop, yeah. Remember the browser wars? Remember the way HTML is? Sometimes sticking to standards, even if they are silly, is a good idea.

2002-08-08T22:58:00.000Z

Fresh Straw

Straw just reached version 0.6. This version reduces breakage quite a bit and adds a couple of new features: show and subscribe to an item's source easily and poll optionally only the selected feed.

2002-08-08T13:04:00.000Z

Greetings from the bottom

Washington Post's story about Argentina's current state is a depressing read. Sometimes things can change fast, and suddenly cheering the collapse of capitalism seems awfully shallow.

2002-08-06T07:08:00.000Z

Treehuggers

So it seems everybody is linking to BlogTree. Dave thinks it's taking off. The question is, how far is it going to go? I'm a bit sceptical about it's longetivity. Things like SixDegrees, IRC galleries, etc, seem to sweep through my circle of acquaintances every now and then: people hear about them, register, make the links, play with the system for a while, then forget it (although in its current incarnation, the finnish IRC gallery seems to be surprisingly successful.)

BlogTree as an idea seems to be kind of interesting, but in the long run, I'm not sure it'll not end up as a warehouse of dead links and irrelevant information, the way these things tend to do. Besides, I have no idea which blogs "inspired" my blog's creation, so I haven't registered it there.

2002-08-01T20:53:00.000Z

Storage

The next logical step for Straw in terms of features would be storage of items locally. Or at least the storage of information about what has been read, but I'll probably go for local storage of feed items because I think it won't be much more difficult.

However, I'm a bit unsure how to go about this. ZODB would be a logical choice in many respects: it would allow me to store relatively painlessly the object structure without me having to do very much at all. Patrick Logan believes transparent persistence isn't usually needed or practical for real problems , but I believe it would suit this case pretty well.

There are two problems with ZODB, however: it apparently isn't very compatible with Python 2.2 — but this might be circumvented with a new enough version and by not using new-style classes, I believe — and it's yet another dependency, this time with a library that pretty much no-one will have installed already and which might require fetching from CVS. I usually don't mind dependencies, they just indicate healthy code reuse, but in this case I'm a bit unsure.

The other option would probably be Sleepycat's Berkeley DB 3. Many/most people already have this and if they don't it's probably easier to install. But I guess it won't be as neat a fit as ZODB, requiring more manual mapping back and forth from my code. Obviously I'd love to avoid that if possible :-)

As you might have noticed, I don't have quite all the facts I need to make an informed decision. ZODB being part of the Zope family, I'm not sure how easy it'll be to correct the situation — the documentation can be sometimes hard to come by. That's likely to be one thing that won't be a problem with DB 3.

Time to find out about the state of the art of this persistency thing.

2002-08-01T19:54:00.000Z

It's release season

Straw is now at version 0.5.

2002-07-31T22:29:00.000Z

New Straw

A new version of Straw, again. Version 0.4 features a code base I'm somewhat satisfied with, some new user interface features for navigating the items, and probably a whole lot of new bugs and regressions.

There's still some ugliness in the code but I don't think anything too major, so I might be able to focus on new features now, unless I detect some major breakage.

2002-07-29T15:52:00.000Z

Any loser could make a better browser than this, right?

An excerpt from an interview with Chris Blizzard:

Q: What's been the hardest thing about working on this project? A: Imagine a four million line code base that is based on C++, uses threads, a hundred shared libraries, and is over a gigabyte in size when built and then point a debugger at it that was designed to debug GNU sed.

2002-07-29T10:40:00.000Z

Gooed Symbian

Chris Double has provided the first reason I'd almost — if it starts working some time in the future, I'll change that to really — like to have a phone running Symbian.

2002-07-28T13:39:00.000Z

So many books, so little capital

On Patrick Logan's blog, I just saw this note about corporations as a remnant of the feudal system, with a link to a review of The Divine Right of Capital by Marjorie Kelly. It sounded very interesting — I go for a lot of that kind of stuff — so I checked out how much it would cost.

On Amazon, it's pretty ok: $17.47 doesn't sound too bad. On Akateeminen, the largest bookstore chain in Finland, not so nice. I'd be quite willing to shell out a few bucks to support local dealers, but the difference between 40,60 € and $17.47 is a bit more than I'm readily willing to stomach, especially in the days of near-parity.

Of course it's not without costs shipping crates of books to Finland and keeping them in a warehouse somewhere, but still, with that kind of prices, it's hard to see how they manage to compete with web stores, especially in the future if and when more people have the means and ability to access those stores and pay for their purchases. Personal customer service sure is nice, but not that nice.

2002-07-23T16:13:00.000Z

Ha, it's not my fault!

The gtkhtml2 segfault bug isn't the only one that has been plaguing Straw — another, not quite as annoying, one is a problem where selecting a row in a list where there was no selection before causes two selection changed signals to be emitted. I've tried to circumvent this a bit but not quite as successfully as I'd like. I wasn't quite sure if this was a problem with my code — it sure didn't look like it — or a library bug. Well, digging into the GNOME Bugzilla gave the answer to that: bug #82344 describes the problem quite nicely.

Too bad no-one has touched that for two months.

BTW, I think I must be linking more to my own old articles than to any other blog. I'm not sure if this is healthy.

2002-07-23T14:52:00.000Z

Straw segfaulting

At the moment, Patrick Logan's RSS (the item with title "What happened to reason and rationality?") is making Straw crash with an segmentation fault.

This seems to be a gtkhtml2 bug, see bug #87490 and bug #87567. Why of all the feeds I read and all the items contained in them this happens with just that one, at least at the moment, I don't know.

2002-07-23T12:07:00.000Z

Straw 0.3

Straw reached version 0.3. Now the UI is a lot nicer, looking a bit more like a news reader and a bit less like a web browser. I also fixed some bugs, made it cope with the nasty world outside a bit better, and refactored it quite a bit.

I didn't fix the Makefile (see my previous entry) but instead made Straw find the files.

BTW: I wouldn't mind if some artistic soul draw an icon for it. Preferably something that doesn't say "Straw". Something like a straw in a glass of juice or something? And look pretty? :-)

2002-07-19T20:44:00.000Z

Oops

Turns out Straw's Makefile is a bit broken at the moment (version 0.2) — it installs the library directory into /usr/local/lib, which isn't in the default Python search path. You should either adjust $PYTHONPATH or move the directory to /usr/local/lib/python2.2/site-packages or somewhere else really along sys.path.

2002-07-19T19:09:00.000Z

OK, what is it?

John Wiseman is doing it. Cristophe Rhodes hypes it. And Dan Barlow.

So what is it about CLIM, suddenly? I thought it was one of those LISPy dinosaur skeletons left behind on the tapes of time, for archaeologists (wow, got that right on the first try) to discover?

2002-07-16T20:36:00.000Z

And another

Another release of Straw, with support for image loading.

2002-07-16T00:28:00.000Z

Release

The first version of Straw with the confidence-inspiring version number 0.1 has been released.

Summary: blog reading for the GNOME 2 desktop.

2002-07-15T22:02:00.000Z

Why oh why

Why doesn't SAX2 provide a way to access the XML declaration when the encoding information is hidden in its attributes and handling foreign documents without knowing their encoding is sometimes impossible? Great to have an API and then having to work around it. It's not as if it was difficult for the bloody parsers to generate those events, if its considered unclean or something to do it as a processing instruction event, why not a declaration event? Reported to a different object if it makes things any better?

And why isn't GtkHTML2 documented at all?

2002-07-06T20:11:00.000Z

Functionality

Jarno blogged a vintage Tim Peters quote about functional languages (oh well, apparently his archives are broken, as per Blogger usual — the article I'm talking about isn't there.) This reminded me of a post in comp.lang.lisp I saw last year, I believe. However, I was unable to find it with Google Groups, so I'll have to resort to paraphrasing. Someone was asking what are functional languages, or something like that, and the reply was:

"There are two kinds of languages, just like relationships: functional and non-functional."

And while I don't do Haskell and I'm no big fan of functional purity, I have to admit I believe there's truth in that. I'm sure (or at least I hope) timbot would agree that popularity contests aren't the best way to measure the usefulness of language features.

2002-07-05T10:19:00.000Z

More Emacs

As a follow-up to my note about using manual indentation in sgml buffers, here's how to make backspace behave the way it should: delete backward if there's a character to delete, unindent otherwise. This time in a code element, too :-)





(defun delete-backward-or-unindent-tab-stop ()

  (interactive)

  (let ((saved-pos (point)))

    (if (save-excursion

          (block 'look-for-ws 

            (beginning-of-line)

            (let ((tmp-pos (point)))

              (when (= saved-pos tmp-pos)

                (return-from 'look-for-ws t))

              (while (< tmp-pos saved-pos)

                (when (not (memq (char-after tmp-pos) '(?\t ?\ )))

                  (return-from 'look-for-ws t))

                (incf tmp-pos)))

            nil))

      (delete-backward-char)

    (backward-move-to-tab-stop)

    (delete-region (point) saved-pos))))



(define-key sgml-mode-map [backspace] 'delete-backward-or-unindent-tab-stop)

This has been tested only minimally.

Makes me wonder, though, how much of this could have been done without coding elisp — I wouldn't be surprised if I found out there was, after all, already support for all this, somewhere.

Whee - putting the code in a pre totally overflows it out of my layout. Oh well.

[UPDATE]

So the first version was broken. I warned you. Minimal testing still applies :-)

2002-07-03T11:11:00.000Z

Respect and a me2

John Wiseman has discovered Automator's hip-hop too. The strange, futuristic albums Dan the Automator has done with a cohort of other people absolutely rock and are definitely worth checking out — they aren't quite mainstream hip-hop (although Gorillaz is, by definition, mainstream something, being as popular as it is) and are all the more interesting because of that.

2002-07-02T14:09:00.000Z

Emacs tip for the day

It's nice when your editor does the indenting of code for you. It's definitely not nice when your editor tries to do the indenting, but screws it up. Sometimes it doesn't have much of a chance; html-mode, psgml-mode or whatever in Emacs is nice for editing Zope DTML documents — yeah I know, I should get on the DPT wagon — but, not surprisingly, it fails pretty miserably with indenting.

So it's time to go for manual indenting instead. For editing things consisting of blocks, like code, Vi-like indentation where you stay on the indentation level you have set until you insert/remove indentation is much nicer than having to indent separately every line. Me being an Emacs boy, however, I'm not going to switch to using GVim, even if it is very nice. So we need to make Emacs behave sanely.

First: tab needs to be bound to tab-to-tab-stop. Easy enough. Second: how to de-indent? Um, er, hey, there's no function to move back a tab stop? This seems to be really well thought out. Google groups help here: Re: tab-to-tab-stop /backwards/. Finally, how to make Emacs stay on the specified indentation level? The functions indent-relatively-maybe and newline-and-indent, plus the variable indent-line-function are your friends here. Something along these lines in .emacs more or less does the job:

(add-hook 'sgml-mode-hook
(function (lambda ()
(setq indent-line-function 'indent-relative-maybe))))

(define-key sgml-mode-map [tab] 'tab-to-tab-stop)
(define-key sgml-mode-map [(shift tab)] 'backward-move-to-tab-stop)
(define-key sgml-mode-map [return] 'newline-and-indent)

This modifies the keyboard mappings in all the sgml-mode buffers, not just those editing DTML. To avoid this, a separate major mode would have to be defined.

Un-indent is bound to shift-tab which is a bit dodgy, too; it really should work with backspace.

2002-07-01T14:05:00.000Z

Hot launchers in drag

I've been bitten a few times by this: I have in my GNOME panel a bunch of launchers. I open a directory in Nautilus, try to drag and drop a file on a launcher, get the application started without any files open. Duh.

I finally got annoyed, tried to edit the properties of a launcher, hit Help, didn't find any mention of this, tried Google, no luck, tried GNOME Bugzilla, and finally got lucky: via #75470, List of valid Exec parameter variables. Hooray.

2002-06-27T23:04:00.000Z

Fruits of my efforts

Lookee, a screenshot. I thought a desktop blog aggregator would be a nice little project to play with when learning gnome2. It's not ready yet for public consumption, but it's getting there. ATM it handles RSS 0.9x – even that weird "0.94" stuff Dave is generating which most certainly is not documented where his docs tag claims it would be. RSS 1.0 isn't yet supported, polling is done manually for the moment, and the appearance of the document can't yet be customized. All this will be added, hopefully next week. And maybe I'll make the sources available soon too, I just want to polish them a bit before that (-;

Watch this space.

2002-06-27T21:07:00.000Z

Language choice

I started looking into developing some desktop stuff with Gtk2/GNOME 2 stuff. I've used the libraries before, but only in the 1.x versions. My language of choice, in this case, is Python (I'll have to check the state of the Goo bindings some day now.) Given that, it's funny how much C source code I've ended up reading in the last two days.

Want to use GtkHTML2? Apparently the only documentation is in tests subdirectory of its source tree. GConf? examples/basic-gconf-app.c. The documentation examples for the new (pretty nice) tree/list widget were a bit lacking, too, but at least there was API documentation (which needs some reformatting even more than CLHS, dunno how hard it'll be but I'll probably have to try sooner or later.)

Someone announced a Python GTK+/GNOME Wiki just a few days ago – after this is done, I'm probably going to document some of the surprises and difficulties (with at least some kind of solutions) there.

2002-06-27T15:17:00.000Z

Meta meta

Old as it appears to be, I only just today came across Metacrap: Putting the torch to seven straw-men of the meta-utopia (thanks to it being in the topic for #python) by Cory Doctorow, also of Boing Boing fame.[1]

Hearing a lot of talk about semantic web (and in in the blogging "community", too,) I agree with Doctorow. Some web people seem to think that getting rid of the font tag makes things suddenly somehow semantic (meaning sweeter, better, cuter and cooler). Other people believe that semantic heaven will come when we embrace customized DTDs, schemas, or whatever is the XMLish buzzword of the day. But inventing fancy tags does not things semantic make, really.

The word semantic implies meaning, and there has to be an agent who understands the meaning the person who inserted the metadata was trying to give to his (non-meta)data. And regardless of how descriptive the tags are and how much there is metadata between them, unless someone has made some pretty radical steps in machine understanding recently, the people hyping the semantic web are going to find themselves in a re-run of the AI Winter if cooler heads won't prevail.

Metadata is useful, but it's unlikely it'll save the world.

[1] Don't TLDs sometime just drive you crazy? This wasn't supposed to point to some design company, the way it did at first.

2002-06-24T09:51:00.000Z

Veggie zope

So I am (wonder of wonders) starting to get the hang of Zope, even my first little product is almost finished. Zope's documentation isn't all it could be - there seems to be a lot of text written, but it's spread all over and often somewhat out of date (and it's not always easy to figure out how out of date it is.) Making things work isn't everything, though - I'm not quite sure if the stuff I'm doing is good style, which is also something I care about and is far harder to figure out, because it varies from context to context. And there's more to good style than aesthetics - bad style may also mean that it's cumbersome or non-consistent to use.

The next step is trying to understand CMF. The same complaint about documentation applies here, except that there seems to be far less of it. And I keep wondering should I use CMF 1.2 as packaged in Debian, or 1.3 beta. And there's Zope 3 coming soon-ish, too.

2002-06-24T07:55:00.000Z

Stock phrase #9

As Dan pointed out, my archives were broken, as per Blogger usual. Of course, it was my own screw-up. They should be ok now (fingers crossed).

2002-06-20T07:36:00.000Z

Adding insult to injury

Stupid Expat/PyExpat/PyExpat SAX2 driver throws fatalerrors and won't go on after a mismatched tag. Ain't this great: first they won't use sexps and instead go for format which requires balanced start/end tags, then the software totally gives up when the tags aren't balanced. So much for trying to help the user.

Argh.

[update]

And PyXML makes it pretty "challenging" trying to figure out what is wrong when a SAX driver won't work. libxmlop, pyxml 0.7.1, python 2.1. Somewhere it's trying to call __init__ with a wrong number of arguments. Clues welcome.

2002-06-19T08:05:00.000Z

Desktop silliness

Browsing the Gnome mailing lists, over at nautilus-list archives I came across this little gem: Unifying ~/Desktop. Ye gods. I was hoping the desktop folks would be, at some point, getting rid of the desktop directory silliness. I have yet to understand what's the point of having a separate home directory and a desktop directory. $HOME is for me to work in, right? So what was the point of having a ~/Desktop, ~/.gnome-desktop, whatever, again? To have a desktop that has nothing to with reality and the stuff I'm working on, I guess.

I'd love to know have the two options, of having a separate desktop and having a $HOME desktop, been tested on users? I don't think it was in the Sun report, I don't know about things Eazel did. Not to mention other environments besides Gnome.

[update]

Oh well, reading a bit further, I found Jeff Waugh testing $HOME as desktop on users. I know the problems he listed (I still haven't managed to convince Gnus to put News and Mail directories somewhere besides $HOME, and ~/evolution is just as stupid), but still, I don't think that's an argument against $HOME as desktop, but against stupid software which clutters your home directory with non-dot directories not meant for the user to use.

2002-06-14T19:14:00.000Z

Usinability

A few notes about XChat's (version 1.8.9, hopefully these have improved with version 1.9 which has apparently a rewritten GUI code for Gtk2) UI, after a day of using it:

To modify the IRC colours, you need to use the Palette dialog. There is no further explanation for the colours, besides the names visible in the window. You can adjust each colour individually, but the changed aren't mirrored by the UI instantly. To see how your change has affected the display, you press the "OK" button, which makes the dialog go away.

The settings dialog has an Apply button, luckily. However, some of the changes will only take effect after you close the IRC window. Confusingly enough, not just the leaves, but also the branches contain configuration options, just like Mozilla.
Some of the options are accompanied by enlightening tooltips. The option "Old-style Nickname Completion" has a tooltip making it all clear: "Nickname completion is old-style (instead of GNU-style)". Of course, not all the options have tooltips, but some of them do have one-line explanations in the manual. On the other hand, the manual also describes some options which can't be found in the UI.

Key bindings are modified via the Edit Key Bindings dialog, with yet another appearance and logic of operation. Without OK/Apply/Cancel, your changes affect the UI as soon as you do them. The lower part of the dialog contains an initially empty field the same colour as your IRC window. If you resize the dialog, weird things happen.
If you press the "Add new" button, a new line is appended at the end of the list, but nothing else happens. The text fields are always active, but they don't have any effect if you don't have a line selected.

The text fields have slightly erratic behaviour: the "Key" field contains always the name of the last key you pressed. If you pressed shift, it says something along the lines of "Shift_L", if you pressed a, it contains "a". This change is instantly visible in the list. The next two text fields are still a mystery if we are adding a new key binding. However, they behave like normal text fields, and things you type into them aren't mirrored by the list before you press enter.

Next up are the modifier selectors: these, again, modify the entry in the list instantly. Of course they don't mirror the list of your real modifier keys, so I don't get to choose Hyper or Super.

Finally there's the option menu. It gives you a list of functions, and when you select one, the explanation for it appears in the strange box underneath. In the same font as you selected for your channel windows (or maybe dialog windows? I don't know, I didn't see a need to use different fonts for them), and if you happened to choose to display time stamps before messages, you get a time stamp in the help box too. However, here you luckily get explanations for the "Data 1" and "Data 2" fields, the meaning of which vary from command to command.

The server window has its share of problems too, but I don't think I'll go into them now.

However, despite the above, I think I prefer XChat to the other currently available options. And maybe some of the problems have been fixed. If not, after the move to Gnome2/Gtk2, I hope I can do something about them. At least I have the source.

2002-06-13T22:44:00.000Z

Easy to use Euros

I've been wondering for a while why I can't get the Microsoft web fonts (one of the truly nice things to come out of Redmond) to display the Euro symbol, as it's pretty obvious they do support it. The problem came even more "acute" as I started to IRC with XChat and dircproxy and decided I wanted to use Georgia as my font.

After Googling and apt-cache searching for a while, I finally found the reason and the solution: the program I've been using to generate fonts.dir files for xfs to read is mkttfdir. I don't remember how I found it, but there it is. It has been doing it's job mostly OK, so I haven't bothered to mess with it, but turns out it handles only ISO-8859-1 and JIS. Now I found out there's another fonts.dir from TTF generating program, called ttmkfdir (naturally!), which does all the encodings. I'd just love to know why the apparently broken mkttfdir is packaged in Debian.

I guess there's a HOWTO or FAQ somewhere which explains this all. I just love it when things are intuitive.

2002-06-13T09:15:00.000Z

me2

Jarno listed many of the problems surrounding moving a blog I've been thinking about too. Luckily I'm too lazy to have a RSS feed without the platform doing it for me, so no problems from that.

2002-06-13T07:35:00.000Z

Patch it up

XML-RPC works via HTTP POSTed XML messages. Araneida, as of version 0.61, assumes all POSTed content is form/urlencoded, but that isn't correct in this case. I have a mostly-working XML-RPC server implementation in Common Lisp ready but I wanted to get it working on Araneida, so this was a problem. I patched Araneida to figure out from Content-type how to handle the content; however, sending just a patch to allow not to modify the content seemed kind of silly, so I ended up doing support for multipart/form-data too.

A few minutes ago I sent the patch to Dan, here's hoping he'll accept it. I kind of expect it to be shot down in flames, as my Common Lisp is still a bit shaky :-) We'll see.

About XML-RPC: it seems to suffer from some of the same problems as RSS apparently does. Weak specification (there should hardly be need for a FAQ to figure out the protocol) and the only way to figure out if you have it more or less correct is to use the UserLand validator. It does have its strengths too: it's extremely simple, but often good enough to do the job.

2002-06-12T07:58:00.000Z

Attack of the Clones

Not surprisingly, I'm not the only one working on a LSP-like thing. And he's quite right in that A) everyone has done one already and B) all this would have been a lot cooler in 1997. Of course, I'll probably do my own thing anyway because the code is already there. ATM it looks like it'll be for Araneida, not AllegroServe, though.

2002-06-08T14:06:00.000Z

In the Navy

So it was finally time for me, too, to do my mandatory civil service of 13 months (well, the other options are 6, 9 or 12 months in the army or 197 days in jail), just like some friends of mine. Last tuesday I came away from the pseudo-educational thing Lapinjärvi, and yesterday I started my work at UIAH's Media Laboratory (where, apparently, Ilya of Suodatin just did his 12 months too.)

So for the next year, I'll be getting better acquainted with Zope, as I'll be working on Fle3. This will be interesting, and hopefully in a positive sense.

2002-06-05T23:50:00.000Z

Maxwell Smart

Here's a smart bookmark (works at least in Galeon) for CLiki lookups: CLiki

2002-06-02T13:00:00.000Z

Comments

Jarno just removed the comment feature from his blog. I've been thinking about this a bit (and I know I'm not the only one, I just saw someone talking about this, too, wish I remember who it was) and I believe Jarno did the right thing: the comment systems undermine the idea of blogs. On John Wiseman's blog, Alex Moffat commented on Paul Graham. But I read Alex's blog occasionally, too, and his comments on the subject (which is quite on-topic for the general theme of his weblog) won't be found there.

I feel that usually if you are commenting on subject such as Graham's papers, it's better to do it on your own blog. If you want to comment on something said on someone else's blog, do it on your blog, make a link and hope that the person you are linking to reads his referrer logs. Ego boosting isn't the only thing they are useful for.

2002-06-02T12:05:00.000Z

Boxy but good

I finally got around to downloading Common Lisp HyperSpec 6.0. It was slightly prettier than the previous version, but still needed some cleaning up for my tastes – I prefer a reference I'm consulting often not to have superfluous images and overlarge formatting. Tidy, Perl (!) and CSS help a bit.

First, I run tidy:



$ tidy -q -w 10000 -asxml -c -m Body/*.htm Front/*.htm Issues/*.htm

tidy gives some errors, but now we have mostly valid XHTML files with
some CSS. Next, we use Perl to modify the documents a bit more. I call
the following reformat.pl:



#!/usr/bin/perl -wpi



s{<img.+?alt="(.+?)".*?>}{$1}mg;

s{(</title>)}

 {$1\n<link rel="stylesheet" 

  href="../style.css" 

  type="text/css" 

  media="screen" />}g;

s{</h1>}{}g;

s{(\QSpec (TM)]</a>\E)}{$1</h1><br />\n}g;

s{\QXANALYS]</a>\E}{Xanalys]</a> }g;



$ ./reformat.pl Body/*.htm Front/*.htm Issues/*.htm

(The regexp substitution which was broken to multiple lines should be all on one line.) After this, just put a stylesheet with the name style.css in the HyperSpec root, something like this:



body { 

       color: black; 

       background: white; 

       margin-left: 3%; 

       margin-right: 3%;

       }

a:hover { color: #aa1050; }

h1 { 

     border-width: 1px; 

     border-color: black black black black; 

     border-style: solid;

     padding: 3px;

     width: 100%;

     background-color: #dddddd;

     font-size: 1.3em;

     }

h2 { font-size: 1.2em; }

And the CLHS experience is so much more pleasant.

(Note: The files in the CLHS tarball are mode 444 by default, so you need to change that before doing any modifications. tidy's -m flag fails silently if it's unable to overwrite the file it's working on.)

2002-06-02T07:52:00.000Z

Lizard

Dave Hyatt doesn't make Gecko sound too good. The scary thing is, he knows what he's talking about.

2002-06-01T08:49:00.000Z

Pen is mightier

There's a new free Lisp book available in PDF form, Loving Lisp, by Mark Watson.

2002-06-01T06:25:00.000Z

Making it easy

I just did a small project with PHP and MySQL.

I used to be a professional Perl programmer. After a few years of that, I had pretty firmly decided that Perl wasn't a language I enjoyed using, thanks to several language design features that struck me as unaesthetic and the lack of a proper REPL.

Perl has been a big influence on the design of PHP, the similarities both in the syntax and in the logic of the language are numerous. This was only my second project with PHP, but I already notice that I don't like this language either.

However, there's one thing PHP is very good at: it makes the implementation of simple web sites very, very easy. I've worked with uncountable templating/embedding/specialtagifying systems during the last seven or eight years and implemented quite a few myself, and I have yet to meet one I'd consider really good. Usually they either lack power of expression (template engines which don't offer access to the underlying "real" programming language) or they aren't easy enough for the non-programmers (embedding real programming languages in the HTML with no easier syntax.) I don't care if non-programmers can't work on the stuff I do on my free time, but when trying to earn my pay that's often a viewpoint I must consider too. PHP seems to strike an adequate balance: it allows only embedding of code, but the programming language is rather simple and resembles many other languages web developers might be at least somewhat familiar with.

Lately I've been working on – among other things – a system for embedding Common Lisp in the HTML templates. It solves one of my needs: I can embed a powerful language I enjoy using in HTML documents. On the other hand, it doesn't allow your average web developer with little coding experience to modify the templates. That might be a problem, at some point, but I've decided to ignore it for now. I don't know if I can make it quite as easy as PHP, and should I even try – maybe something with a bit more structure wouldn't be a bad idea? And I'm still not quite sure should I build on top of Araneida or a servlet-like package I've been working on which uses mod_lisp (nice URL.) I guess I'll have to take a better look at Araneida to figure out if it's suitable. We'll see if I ever get this finished: reading the embedded code is the easy part, trying to make the framework sensible is more difficult :-)

BTW, the combination of Emacs 21 + mmm-mode (after making it recognize php-mode as a cc-mode derivate) + php-mode + psgml + css-mode made the editing of HTML with PHP and CSS almost nice. Almost. Still waiting for the day I'll be using an editor that doesn't need kludges like mmm-mode which really can't hide the brokenness of the system it's been built on.

2002-05-29T13:50:00.000Z

Changing colours

[via mpt's blog] Bowie J. Poag is talking about System 26 (whatever it is, apparently yet another themes.org replacement) and telling "It'll be a day or so, but, support for AfterStep, Blackbox, E, Fluxbox, FVWM, IceWM, KDE, Sawfish and WindowMaker is right around the corner. And no, we won't be supporting GNOME. Its time for the Linux community to rally around one standard, and KDE team has proven their right to be that standard."

Not to mention the silliness of enumerating nearly a dozen window managers and then saying that the community should rally around a standard desktop environment, wasn't Bowie one of the geniuses tearing apart the gnome-gui list back in 1998, trying to, with very little success, to produce a GUI standard or something like that?

I wonder what changed his mind.

2002-05-26T13:23:00.000Z

No more wall

There's an interview with Stefan Seefeld of Berlin fame at Advogato.

Berlin is one of those projects which have been going on for ages, and still everyone's waiting for the thing get ready. I seem to remember it getting started as a "X11 sucks coz it's so slow" thing, in the dark ages of the mid-90s. It has luckily changed direction since then and from a technical stand point looks quite promising, although I'm not very much in love in CORBA.

The interview leaves out some interesting questions: do they have the momentum needed to make it usable? How many active developers do they have? Changing the name of the project seems to be one of those meaningless gestures of meta-activity that does nothing for the project. And Berlin is a cooler name than Fresco, anyway :-) But it's their call.

I wish them best of luck, and hope that someday their software will be a serious alternative to X11 (assuming we're not all in Aqualand by then.)

2002-05-26T12:03:00.000Z

Hello, world - trying out blogger.el now...

2002-05-26T11:45:00.000Z

All together now

Happy birthday to Jii!

2002-05-26T10:20:00.000Z

Missing links

Anita Roddick's anecdotes about censored (well, not in the strict sense) ads are quite funny, in a sad way. However, this thing about Malkovich was a bit annoying: Roddick didn't give any links to the story about Malkovich wanting to shoot Fisk, only to Fisk complaining (for a good reason) about it. And of course, Fisk's story was from a newspaper, so there were no links there, either. Google to the rescue.

I wonder how long will it take for some subset of the major newspapers / news agencies to provide links in their articles. Three years? Ten? Most approach the web (or internet, if you will, no need to limit this just to HTML browsers) in an awfully newspaperish way, exploiting none of the possibilities, with the occasional exception of constant updates. The main reason seems to be inertia.

Helsingin Sanomat is probably the only halfway decent newspaper in Finland. They have a rather nice web presence, which they advertise too, saying it's one of the things you get when you subscribe to the paper. Only there's no need to subscribe, it's available for free for anyone. Oh yeah, and they recently scaled down some of their web operations, because it wasn't generating income. I think there's no need to go on.

2002-05-26T08:15:00.000Z

Paul Graham and the Art of Language Wars

Everybody is blogging Paul Graham. Graham has written a lot lately about the relative power and succintness of various languages, mainly Python and Lisp. This is very much a continuation of his work on Arc (which seemed somewhat interesting, but flawed, from the first reading of the paper), a new dialect of Lisp.

I can see why he designed Arc; Common Lisp has its share of problems, and its progress has pretty much stalled. However, these papers about succintness look to me like a high-brow, pseudoacademic variant of mega-crossposted "my language is better than yours" usenet trolls, however much he tries to appear objective.

Everyone has the right to express their opinion, so I'm not going to question his reasons for writing those papers. But I'm going to state my opinion about their relevance: they are not going to make anyone switch languages, and their impact on people designing new languages will be quite small, too - when someone starts to design a programming language, for whatever reason, he will have his own ideas about what makes a good language and reading Graham's papers will not convince him to make it a Lisp (or less verbose than Python if he likes the language.)

2002-05-20T13:54:00.000Z

Make world

More things I really, really need to write myself: blogging software. Blogger's (and BlogSpot's, maybe I should pay the price to get rid of the ads) idea of nice HTML isn't quite the same as mine. And it's pretty limited in other ways too - I was surprised it doesn't support automatic RSS creation.

Apparently Marko is doing it too (he calls it Blogiston), but I suspect his ideas of a nice system differ from mine somewhat (not going to work on anything besides a Mac, for instance.)

2002-05-20T09:15:00.000Z

Being public

When you start to blog, you have to make some decisions. At least I know I have to. Should I tell the whole wide world of Internet everything about my personal life? Should I start playing a pundit? Should I post links to every vaguely entertaining page I come across, with a one-line comment, or should I create more content myself? Should I play a smart-ass, or keep the tone more formal? Maybe for some people these decisions come more easily - I suppose often the decision to start a blog comes from a need that dictates the mode of writing, too. Maybe it's so for me too, I just had to first decide why, exactly, I wanted a blog.

I decided the blog has to fill some space that is not already covered by my current methods of communication - I want to write about ideas that are easier to express in writing than in face-to-face discussions and write about things that I feel need more permanence than a one-line expression of thought on IRC. Areas to cover will be mostly technical, stuff like free software and software development, relating to things I do day to day (well, otherwise it would be pretty pointless to call it a blog.) I could use news, but I don't much feel like participating in flame wars. And I don't read usenet that often these days, anyway.

And because I'm a non-native english speaker, I had to make the decision of language, too: my Finnish is far more fluent than my English, and that goes for most of my friends, too. On the other hand, the things that I intend to cover here would have a rather limited Finnish-speaking audience. So English it is. Maybe my writing will improve along the way, too.

The next thing to do is to actually blog about something besides blogging, I guess.