Category: Technology

So, ESR always talks about how fetchmail is the best UNIX program on the planet: well-written, standards-compliant, stable, etc.

That’s why I use it (with procmail) instead of downloading messages with something like evolution or thunderbird. Basically, evolution isn’t standards compliant and so it doesn’t deal with my mail server well, and thunderbird is ugly.

But evolution does let me use normal mail spools, which fetchmail and procmail can deliver to.

But fetchmail is annoying in some ways. For one, it stores passwords in clear text, in the /etc directory, and ESR thinks that’s good enough. If you don’t want to run daemon mode (like me), since I want to be able to quickly download the latest e-mail, that means plain text passwords in my home directory. I’ve written a shell script fetchmail wrapper that gets around this restriction so my password is never stored on my hard disk at all, but only passed from my keys to fetchmail directly, but what a pain, really (I wrote this a few weeks ago).

Now the other thing is that i have four mail accounts and it serializes access! Jesus! It doesn’t open up connections to each of my accounts, stick messages in some buffer and then write them to disk serially. Instead, it does wait-request-wait-request for each message of each account in order, and then at the end of a message download it actually delivers the message to procmail. So that means it takes me about 30 seconds to check mail on my four accounts, whereas it could take as long as my slowest account (pixelmonkey), which is about 10 seconds.

So, I just found it funny that someone got so fed up with this that they wrote retchmail. Here’s a quote, and a link:

quote:

---

Eric S. Raymond’s Fetchmail is a full-featured, robust, well-documented remote-mail retrieval and forwarding utility intended to be used over on-demand TCP/IP links (such as SLIP or PPP connections). It supports every remote-mail protocol now in use on the Internet: POP2, POP3, RPOP, APOP, KPOP, all flavors of IMAP, ETRN, and ODMR. It can even support IPv6? and IPSEC.

Oh, you were asking about RETCHMAIL. Sorry.

RetchMail is the world’s most stupidly fast POP3 retriever.

It is almost completely unlike fetchmail: RetchMail lacks features, isn’t particularly robust (although it won’t delete a message until sendmail says it was delivered okay), has nearly no documentation, and is actually fast.
…
5. Is RetchMail standards-compliant?

Quick answer: HA HA HA HA HA HA no.

Just to be extra clear about this: yes, it is morally wrong for us to have written RetchMail, and it is morally wrong for you to use it. But try it, it’s really fast!

---

Heh, that’s open source; it’s all about choice. Here’s a link.

I read this on a Gentoo forum today about Alsa’s DMIX plugin:

quote:

---

I mean, why can’t they make a _transparent_ software mixer? Why must every software be “updated” and “configured” to support the software mixing thing? Why on Earth couldn’t they make a plugin that needs no attention from the actual software, but just lets the software open and open the sound device? Why must mplayer, arts and others _know_ about the dmix plugin and _select_ it? That is not transparent.

---

The guy is right on. This is one of Linux’s main problems. Although we have as a hacker tenet to never reinvent the wheel, we also have a shitload of libraries and ways of doing things, and sometimes it leads to this awful situation: where you want to get software mixing of audio to work, but it will only work if you enable a mixing plugin in each application.

At the library for yet another full day, reading operating systems papers. This time, however, I kept getting really bored (can you blame me?), and I kept thinking about so many other things. They take so long to read, so much attention… I have so little to give. I don’t even have my laptop here, and I only spent the last 30 minutes on a computer.

This is my “break,” searching for vim plugins:

ZoomWin for maximizing a buffer temporarily, MiniBufferExplorer for switching buffers (Matt would probably like that one), Project Explorer for “favorite files”, ShowMarks for finding marks you set in a file, SearchComplete’s tab completion for searching (s/bla/bla2/g), cppcomplete for code completion (Matt says it rocks, like intellisense for C/C++/Java)…

I think I’ll actually install most of those (provided they work as advertised).

I think for C# programming all one needs to do is modify JavaBrowser and cppcomplete the recognize all of C#’s syntax, and we’d be all set C#-wise. Hell, maybe I’ll even do it next month.

While programming today, I came up with some nice vim bindings. If you use vim, and want tags-powered “intellisense-like” stuff, use this. Also thrown in there are hotkeys for :Tlist (indispensible vim plugin for ctags integration) and saving quickly (I save often, so I like to have a single key in either normal or insert mode).

" intellisense - yea well close enough
noremap <F2> mz<C-w>1]
noremap <F3> <C-w>c`za
noremap <F4> <C-w>11_

" for insert mode, goes back to unmatched brace
" so it's good for function prototype lookups
inoremap <F2> <ESC>mz 999[(b<C-w>1]

" tlist hotkeys
noremap <F8> :Tlist<CR>
noremap <F9> :TlistUpdate<CR>

" save hotkey for normal and insert modes
noremap <F10> :w<CR>
inoremap <F10> <ESC>:w<CR>a

Update from the future: The remnants of this 2004 ~/.vimrc customization are still present in my $HOME dotfiles, many years later. You can find my recent .vimrc on GitHub here.

I spent about 5 hours today doing algorithm homework/studying. I like that class a little more now, even though some of it is a pain. Divide and conquer and dynamic programming actually are powerful concepts, once you get a feel for them.

Then I came back to my dorm, ate some food, and played with vim for literally 2 hours. I guess that was a waste of time, but I learned so much about this editor. Now I feel I can be twice as productive when I code. Especially with all the stuff I implemented for prototype previewing in my vimrc, and with all the support for ctags vim always had but I never used. Wow, this is one powerful programmer’s editor.

Matt has been telling me that I should use emacs with vi emulation (viper), and then I’ll get access to all that great emacs stuff. Maybe. I have nothing against it, except that emacs seems like an operating system unto itself. Eh, it doesn’t matter I guess, I just need to know one of the two well. They are both portable, and run on all major platforms.

Computer scientists have a lot to learn and realize.

For one thing, computers aren’t the center of the universe. What may be an ideal for computer scientists may not be an ideal for normal people. And very often, computer scientists affect normal people because everyone uses computers (or at least, everyone will).

All these computers sit around on our desks, we only use a fraction of their power at any time. Right now, I am typing this blog entry, using less than 1% of my CPU’s power. Theoretically, it could be doing things–helpful things, things that will make my life easier. It could be doing smart analyses of what I’m writing and try to predict what I’m going to do next. It could be some sort of extension of my mind, helping me produce better work. Instead, it sits there idle, useless. Glorified typewriter.

There have been many innovations, but the “humanity” of computers has been lost. We shouldn’t be designing our lives around them… they should be designed around US. The way the human mind works needs to be complimented. I should not change my ways for the computer, except if in changing them it makes my life easier, less complicated, and makes me more powerful as a human being.

These should be enabling devices. They surely have the potential to be enabling devices. But right now, in many ways, they disable us. We are restricted by the rules programmers place on us. We live under a sort of “law of code” which Lawrence Lessig describes in his books. In addition, the gov’t and other groups seek ways to use computers to control people.

Computers need to become more like us, so that they can seem familiar, useful, but at the same time, a whole lot more powerful. What computers have that we don’t have is speed, time (CPU time), and infinite storage. What we have that computers don’t have is the ability to reason about our experiences in very flexible ways. Wrapping a computer’s speed and storage capabilities around our own flexible abilities as conscious beings would mean a very powerful harmony.

Why are we still talking about how to isolate faults, or make device driver subsystems better, when this human element is so sorely needed, and would be so well appreciated?

I just found this essay on UI design and am surprised I never read it before. Some of the stuff Havoc says rings particularly true, and also confirms the need for some of the ideas I have for Glade-3 (ideas which, I think, will push UI design into the user’s hands, and exploit the end-to-end principal in the world of graphical software development; this is only one of the key points–my ideas also have other implications that, I think, are good for Linux applications in general).

In my spare time, lately, I’ve been drafting out these ideas, and I think soon I’ll have them clear enough to send a “brainstorm announcement” to the glade-devel mailing list. God I wish I had more time for this stuff!

I’m working on this philosophy paper, and am having a bit of a brain struggle. The paper I read makes a very strong point for a model of consciousness that is computational (functional), so that, for example, it is conceivable that a sufficiently advanced computer (or group of computers) could replace the brain and serve the same role (i.e. I’d still have conscious experiences, etc.)… but this is very hard for me to accept at a “gut-reaction” level. Although it would be very easy for me to write a paper defending Dennett’s claim, I am going to have work through this to figure out what is wrong with it (I am convinced something is wrong with it).

A quote Dennett cited (attributed to Fodor) made me laugh out loud and receive stares in this quiet lounge: “If, in short, there is a community of computers living in my head, there had also better be somebody who is in charge; and, by God, it had better be me.”

I have to re-read, and re-read, and outline, and re-read, and maybe, eventually, write.

I’ve been playing with a preprocessor called gob2, which I only recently discovered. I found it because I actually had the same exact idea as the author. I was reading the Gnome 2 Developer’s Guide, and realized that although GObjects are nice and handy when they are already coded (as in all the widgets in GTK), they are actually a pain in the ass to write from scratch. Tons of boilerplate code, tons of macros you need to code to keep yourself sane.

So, I thought, what if some code produced all this boilerplate code for me? Someone already had the idea. It’s called GOB (or gob2), the GObject builder. Check it out.

What’s so nice about it is that it tries to have the feel of Java or C#. So, check out this definition of a new GtkWidget which counts the number of clicks it receives. If you run gob2 gtk-button.gob, it will produce gtk-button.{c,h} with all the proper GObject boilerplate code you’ve come to expect coding by hand.

Much nicer than doing that from scratch.

What’s strange is that more GTK+ developers aren’t using this. On the gob2 mailing list, the author claims it is used in gnome-vfs, but doesn’t it seem like it should be used by a lot of developers? (Granted, it’s not as nice as using gtkmm/C++ I guess, but there are still people declaring objects in C out there).