23 November 2005

MJ Ray implied that anyone who thought the DFSG only applied to programs, not documentation, was negligent about DFSG and that the only such people who are exempt from that criticism are those who joined Debian before the Social Contract. That is blatantly wrong and bordering on outrageous revisionism. I joined Debian two-three years after the Social Contract was adopted. I read it and understood it (at least thought I did) when I joined. However, there was no "documentation is software and therefore must fall under DFSG" movement back then. It seemed to me then, and I still believe it, that the opposite was a silent consensus. I one uploaded into main a package containing almost nothing but explicitly non-free documents. (The documents have since been freed.) I was sincere in that attempt and in my argumentation at that time, and at least one of the then-current ftpmaster team must have agreed with me, since the package passed NEW checks. (Granted, I also argued that the packages would be later freed, but that should not have affected the decision if nonfree documents really had been deemed to be against the SC.) It seems to me that the current consensus is a recent one, certainly less than four years old. I have decided to honor that consensus, but don't come tell me that I was dishonest when I joined!

Date: Mon, 17 Apr 2000 03:07:13 +0300
From: Antti-Juhani Kaijanaho <gaia@iki.fi>
To: "Debian developers' misc list" <debian-devel@lists.debian.org>
Subject: RFC/ITP: Constitution of Finland

On IRC I proposed packaging the Constitution of Finland.  Among other
things, it would be an example of a good Constitution :-)  The response
on IRC was positive; what do people think here?

Since Finnish laws are specifically not protected by copyright (see
Finnish copyright law), the law is free and can go to main.  There is
an English translation at http://www.om.fi/perustuslaki/3313.htm.
I am researching its copyright status.

Assuming people are okay with this, and the copyright issues around the
English translation are resolved, I will package Constitution of Finland
to woody.

-- 
%%% Antti-Juhani Kaijanaho % gaia@iki.fi % http://www.iki.fi/gaia/ %%%

                    I'm moving IRL on May 2, 2000.
               New contact information on the home page

Do you remember this infamous thread? :) For some reason, it has almost completely been censored off the Debian mailing list archives. I digged my own archive and put the thread online in a mailbox. Have fun with nostalgia and AJK-bashing :) Maybe this will help to reestablish the thread in the archives; who knows?

Jordi Mallach expressed fear, uncertainty and doubt about the European Union Consitution. I would like to make the following two points:

1. Much of what is in the proposed Constitution is already in force as part of the current foundational treaties. So, if you read the Constitution and want to bash it, please read also the current treaties. I expect you to find it to be clearer and content-wise quite similar to them. So don't think that this document is an evil attempt to seize power or something (well, it might be that but I doubt it).
2. A change to the current foundational treaties would be necessary in any case. The acceptance of new member states makes decision-making by the current rules difficult, if not impossible.

I assume most of the FUD spring from the grand name of the new document, a Constitution. It deserves the name quite rightly. It is the sole foundational document after it is accepted, replacing quite a few treaties. However, its impact is mostly illusory. A similar change was done in Finland a few years ago, when the three constitutional statutes (none of them called the Constitution) were combined and updated into a single law called the Constitution. (Long-time Debian developers probably remember my ill-conceived ITP from four years ago:) Perhaps this FUD is exactly why the consitution should not be subjected to referenda. I don't doubt that Jordi is not alone in his feelings, and such feelings could easily cause it to fail in a referendum. The only effect of such a failure would be making decision-making in the EU painful, without actually addressing the issues people have, such as Jordi's regional issues.

Can we agree on not calling Anthony Towns AJ? Every time someone refers to AJ in the lists or in Planet Debian, it takes me a few moments to realize it's not me they are talking about. Let's face it, AJ is a fairly obvious nickname for anybody with the (partial) initials A. J. And there are bound to be many of those. I personally discourage people calling me A(-)J, but there still are many people who do so. Since I have been called that by many people since eigth grade (around 1991), I have learned to respond to that name. I personally use it only in the three-letter signature AJK. It took me some time to realize that Joe Wreschnig wasn't calling my reading of the old SC very strange, and even now I cannot be 100 % sure. Update: Joe has updated his post to no longer refer to Anthony as AJ.

If you have a firm opinion about the latest general resolution vote, please publish your ballot, and an explanation of your reasoning, in your blog (and hence in Planet Debian). Note that the ballots will all be published after the vote, so there is no voter confidentiality protecting your ballot in any case, and hearing your arguments will help people like me, who are undecided, to decide. I'm not looking to imitate anyone in my voting behaviour, but after the previous vote, I don't trust my personal reading of the proposals to be accurate or at least to reflect the reading that matters. Julien Danjou already explained his position:

Since I think we have to release Sarge as soon as possible in order to satisfay our users, I have just voted for postponing changes until Sarge is released.

Julien, I would like to ask you to publish and explain your full ballot. There are a lot of options in there, and it is not at all clear, at least to me, how one would order them to bring about a release of Sarge as soon as possible (note that I am as of yet undecided on even whether I want that or not:).

In some EU countries the polls are still open. Remember to use your voting right responsibly. Software patents in Europe can still be thwarted if the new European Parliament is well chosen. Finnish people, if you haven't voted yet, you are almost out of time. Run, don't walk, to your polling place! I recommend Matti Wuori (101) and other Greens (82-101).

It is annoying to receive hundreds of bounces of mails I never sent. I can sympathise with the senders, although there is an effective technique to cutting them (reject mail, do not bounce it). However, when I get a mail asking me to confirm that a mail that I never sent is not spam - when it actually is - I get mad. It is very inconsiderate to send such mails: it is comparable with emptying your garbage bin on your neighbour's property. People who set up such autoresponders probably do get a spam-free zone; however, they do it by offloading the problem on Somebody Else. It is well known that spam return addresses are bogus, sometimes nonexistent but often owned by third parties who have nothing to do with the spam. I get a lot of spam myself, thank you; I do not need yours.

Remember to vote in the European Parliament elections!

19:30 @<mudyc> A one-question geek test. If you get the joke, you're a geek: 
19:30 @<mudyc> Seen on a California license plate on a VW Beetle:
19:31 @<mudyc> FEATURE

QVOD DEVS OPTIME VERTAT
EX LEGIBVS VNIVERSITAS JYV SKYL ENSIS ATQVE EX DECRETO
AMPLISSIMAE INFORMATICARVM IN EA VNIVERSITATE FACVLTATIS
ABSOLVTO OMNI SPECIMINE LEGITIMO PHILOSOPHIAE MAGISTER CREATVR OMNIBVSQVE HVIVS DIGNITATIS IVRIBVS PRIVILEGIISQVE ORNATVR
IN VBRE JYV SKYL A. D.
XI KALENDAS IVNIAS ANNO MMIV
VIR PRAECLARISSIMVS ANTTI-JUHANI KAIJANAHO PHILOSOPHIAE CANDIDATVS
QVOD FACTVM
NOMINE MEO SVBSCRIPTO ET SIGILLO VNIVERSITATIS ADFIXO
EGO CONFIRMO
LEGITIME CONSTITVTVS PROMOTOR JAAKKO LEHTONEN

---

This is what I got today in mail. I had participated in absentia in the University's promotion, which happens every four or five years or so. The above is the text of my new diploma. I am now really philosophiae magister.

I was playing around with BNFC's C++ mode, and discovered that its authors have fallen prey to two of C++'s famous pitfalls:

• All polymorphic classes (i.e. classes with at least one virtual function) must have a virtual destructor. Fortunately, G++ -Wall warns about this one.
• The compiler inserts a default copy constructor and a default copy assignment operator to all classes that lack an explicit copy constructor and an explicit copy assignment operator. This will bite you if you have pointer members in the class, and your destructor performs a delete operation on one of them. You need to either forbid copying by adding an explicit private copy constructor and an explicit private copy assignment operator to the class (prototypes without definitions will do nicely), or by enforcing some sane copying policy by giving your own copy constructor and copy assignment operator (popular policies are deep copying and sharing with reference counting).

Were you aware of these issues? :) I fixed these problems in BNFC with this patch.

Patriot Act [...] requires John Ashcroft's written permission before you can open your mouth Source: Some Notes on the "Who wrote Linux" Kerfuffle, Release 1.1

I wrote a couple of days ago about safety in programming languages. A certain comment reminded me that people (even programmers) misinterpret technical terms from the area of programming languages. It reminds me of that old joke about the C programmer who is very hurt when somebody claims that C is not a functional programming language - of course it is functional, our hero says, as it functions perfectly. It also reminds me of a quote Christopher Strachey has been said to have written in the 1960's:

It has long been my personal view that the separation of practical and theoretical work is artificial and injurious. Much of the practical work done in computing, both in software and in hardware design, is unsound and clumsy because the people who do it have not any clear understanding of the fundamental design principles of their work. Most of the abstract mathematical and theoretical work is sterile because it has no point of contact with real computing.

In programming languages jargon, safety is the property of a language that all well-formed programs have a well-defined semantics. That is, in a safe language, programs that the compiler does not reject, do not go beserk (corrupt the stack or heap or otherwise invoke nasal demons). A well-defined semantics might be a forcible halt of the program, which is often not the ideal way of dealing with problems (sometimes it is even harmful), but it is beyond the scope of safety to specify a programmer-friendly semantics for such occurrences. Safety is the minimum requirement, not by any means sufficient for all contingencies. There is a related slogan attributed to Robin Milner:

Well-typed programs do not go wrong.

(Wrong, in this case, is an abstraction denoting nasal demons, heap corruption and other such nasty things.)

I wrote two days ago:

Is there a mechanism to figure out what exception it is that fails to be caught by any exception handler on the stack? Of course, it can be caught by using catch (...), but that just tells me as much as G++'s default handler: that some exception was thrown.

Matus Telgarsky later pointed out to me in private mail that G++ does have a better terminate() handler, used by default in G++ 3.4. I could also invoke it in older G++ by adding

std::set_terminate( __gnu_cxx::__verbose_terminate_handler );

which I did. Based on the information it gives I eventually tracked the problem to my use of safe_ptr<AbSyn::AST>::release in my safe_ptr-downcasting functions in ast.hh. I had written that code originally using std::auto_ptr but due to problems I then implemented a reference-counted replacement for it. I had tried to make safe_ptr's interface to look as similar to std::auto_ptr's interface as possible, which now turned out to be a mistake. It turned out that there were four active aliases for things I passed to release at the time, which quite appropriately triggered an exception. However, unfortunately, I had given the downcasting functions exception specifications not listing that particular exception. Now, the semantics of an exception not listed being thrown in such a situation is that the program is immediately terminated. This problem highlights a C++ exception misfeature: it allows calling a function without an exception specification (which essentially means that it might throw anything) from a function with an exception specification (where throwing an exception not listed invokes immediate termination) without so much as a warning to guide the programmer!

Recent discussions within Debian that make a noticeable detachment from the FSF point of view toward freeness, mainly about the GNU Free Documentation License and other documentation licenses, have induced me to wonder whether this is a symptom of a deeper difference of opinion between Debian and FSF. FSF is supposed to be developing the next generation GPL. One of the rumours about potential changes involve a requirement to post sources of GPL'd software that is used to power a website. If GPLv3 really included such a requirement, then I believe it would fail the DFSG (no restrictions on usage). It is also possible that the same difference of opinion between Debian and FSF that is now manifesting itself in the GFDL debate would end up causing us to reject GPLv3, even if it didn't include that particular restriction. It is likely that many free software authors, certainly all FSF projects, would eventually start licensing their software under GPLv3 (or, at your option, any later version). Then we would be in real deep shit. Update 2004-05-18: I seem to have been a little sloppy writing the above. I do not claim that a GPLv3 of the kind I indicate above necessarily fails the DFSG (such a claim would be nonsensical before an actual draft of GPLv3 is published). I intended to say that I expect to see such arguments if and when such a GPL is released, and I would not be surprised if such an attitude were to win the eventual argument. The point of this post is to show where that would lead us. I'll reprint here a comment I posted to this entry earlier:

I have no particular wish to see Debian and FSF disagree about the freeness of GPL, for the reasons I gave in the post.

Martin-Éric Racine proposes development of a new spell checker for Fennic languages (mainly Finnish and Estonian). Since he forbids use of his contact addresses from anybody who is not his customer, I have to write this in public. I was involved in 1998 in a project with similar aims, Oikoluku-SF. It never accomplished anything. My personal belief is that the main problem is getting enough linguistic data to encode the morphology of Fennic languages. All such existing data is to my knowledge non-free. The rest is just a matter of programming: the algorithms are not rocket science.

Is there a mechanism to figure out what exception it is that fails to be caught by any exception handler on the stack? Of course, it can be caught by using catch (...), but that just tells me as much as G++'s default handler: that some exception was thrown. I've been implementing the Hao compiler in C++, as C was too low-level but I wanted something widespread. A public darcs repository is available for curious bystanders to stare at; I doubt anybody gets much out of it (the haoc/ subdirectory contains the compiler in development; the other directories are more or less stale). I push to that repository occasionally; it's not necessarily up to date.

Here is my hackergotchi:

Most programming languages designed before 1980's were safe. All well-formed programs had well-defined meaning. The main exceptions were very low-level languages where run-time tests for indexes and such things were just too expensive. From this low-level tradition came C. Then for some very strange reason C became popular for general-purpose programming. It has been in to avoid safety as much as possible ever since then. For most programming tasks, safety is much more important than speed. It seems it's only now that people are realizing it. If we hadn't made the C mistake, we wouldn't need to be paying for it with all those buffer overflows. Cyclone looks like a useful C replacement for most projects. Never tried it for real, though.

I ran the other day into Volume 13, Issue 1-2 (2000) of Higher-Order and Symbolic Computation, which contains articles by and about Christopher Strachey. From him come terms like l-value, parametric polymorphism, first-class object. He is a classic in the theory of programming languages. The issue is freely available at Kluwer's site. I am unable to give a direct link.