Disqus - Latest Comments for madalu

Re: Why The World is Ready for Dexy

madalu — Wed, 12 Jan 2011 07:37:38 -0000

Great post! Are plain text markup systems used much for technical documentation? Markdown, rest, org, etc. under version control?

I've been intrigued by literate programming recently, i.e., writing source code and documentation in the same file and then "exporting" the final results to separate files. But the effort of keeping things organized can quickly become overwhelming. And the logical flow of code often does not correspond to the logical flow of use and documentation.

I really like the conventions of elisp: functions and variables contain their own "docstrings", which are easily accessible within the Emacs editing environment. (Python has similar conventions, but it's harder to read the documentation when running a Python program.)

Re: Sys Admin Legacies in Free Software

madalu — Fri, 10 Dec 2010 18:58:08 -0000

I think you've highlighted something extremely important about Linux/BSD. Both, in the end, are sysadmin OS's. That is, to use them efficiently, you need to think like a sysadmin. This is a much bemoaned fact among Linux detractors: Linux forces you to be your own sysadmin. But isn't that the whole point?

I mean the following as a compliment: Linux/Unix will always be mediocre as a desktop. That is simply not what it was designed for, and that's the reason that full blown desktop environments often seem like tacked on afterthoughts. It takes heavy distro-specific hacking and bloat to create a "seamless" desktop environment, and many of the things meant to help the end user make Linux more complicated for the sysadmin. To me, Linux is not a desktop, but rather a remarkable set of free computing tools and libraries---the most powerful free toolbox in the world, if you will.

Re: Why Email Still Matters

madalu — Thu, 11 Nov 2010 20:23:09 -0000

Great post! I, too, will cling to email (and resist Facebook) with my dying breath.

I fear that outside of mailing lists, the campaign against top-posting is a lost cause. Most GUI clients make it relatively difficult to quote emails efficiently. I don't know a single person outside of the hacker world who types responses beneath quoted material. And yet this one little trick would solve so many of the human problems you mentioned --- verbosity, no clear action or response, etc. In fact, I think Facebook's wall and Twitter's replies would have seemed a lot less innovative if people had used email clients with proper threading and with easy ways to bottom-post.

That said, one should never underestimate vanity as a driving force of technological change. Compared to the exhibitionist thrills afforded by Facebook, Twitter, blog comments, etc., private email is downright boring. ;)

Re: Phone Torched

madalu — Fri, 22 Oct 2010 22:42:10 -0000

I'm holding out for the next generation of the Nokia N900 - if it or a good meego phone ever arrives. (Or I'm waiting for the N900 to drop below $250). I suppose that I can't get around my insistence on having a fully hackable unix userspace on my phone.

So, for the time being, I pretend it's 2005 and rejoice at being able to buy dirt-cheap PCs, to install insanely great free software, to use a perfectly customized Arch install, and to get a cheap, unsubsidized "dumb" phone that would have required a contract a few years back.

Re: http://benjaminste.in/post/1223476561

madalu — Wed, 06 Oct 2010 22:12:00 -0000

The alpha geeks are no longer interested in their brand of laptop. They are too busy exploiting all the possibilities that big data offers for ruling the world. :)

Perhaps it is a bit misleading to separate maker technology from consumer technology. For today's geeks, maker technology is as exciting as the personal computer and internet were in the mid 90s --- all the fiddling around with arduino and makerbot and whatnot gives insight into the consumer technology of 5-10 years from now.

Another way to put this: alpha geeks recognize that future consumer technology is by no means limited to PCs and mobile phones/tablets.

Unix arguably remains the premier alpha geek technology. True alpha geeks turned to Apple in the early 2000s because it was a pretty and stable Unix, not because it was Apple. Non-geeks flock to Apple because it is Apple, Inc. But the alpha geeks are getting restless...

Re: Console Use for the Uninitiated

madalu — Sat, 28 Aug 2010 08:28:10 -0000

Reading man pages or typing "command-x --help" can make your head swim, but the command line, like language, packs a lot of power into a little space. Perhaps the command line still seems superior because of the persistence of poorly designed GUIs. With the ascendance of multitouch and gesture-based computing, we may eventually come to see the command line as a crude relic.

Traditional destkop GUIs, however, make computers easier to use by dissipating their power across dozens of menus. On the command line you can recombine flags and pipe commands with ruthless efficiency. In traditional GUI interfaces you need a separate menu item or button or cumbersome "Preferences" dialog to express these possibilities.

That said, I've been much taken by Jaron Lanier's criticism of the "lock-in" of the UNIX command line, which, he argues, assumes that "reality" consists of "a bunch of really fast typists." Perhaps it's because I fancy myself a really fast typist that the basic *nix assumptions flatter me. :)

Re: Saved Searches and Notmuch Organization

madalu — Mon, 02 Aug 2010 21:39:24 -0000

I think org-mode stikes a nice balance between structure and search. Its outlines can be as organized or disorganized as one likes. At the same time, each outline heading/tree serves as a node of information containing metadata and (optionally) other nodes that can be queried in all sorts of helpful ways.

Re: In Favor of Simple Software

madalu — Tue, 13 Jul 2010 21:43:11 -0000

I think you've identified the key here. Great (is it silly to say zen-like?) software is porous: it 1) runs quietly in the background, 2) is easily accessible by the rest of the system, and 3) can be extended in all sorts of ways. My favorite recent piece of software is mpd (music player daemon), an insanely elegant music library solution that can be accessed in all sorts of ways.

In addition to problems of quality control, there is also the problem of packaging. The commercial software world understandably prefers walled gardens -- i.e., self-contained applications, or "apps," as today's silly mobile nomenclature would have it.

Linux/Unix is less a room full of finished furniture than a powerful workshop full of wonderful tools. Understandably, not everyone wants to spend all their time in the workshop.

I'd like to hear more about the benefits of irssi. Right now I'm using the very simple rcirc (included with emacs), so all my irc channels occupy separate buffers.

Re: Organize Your Thoughts More Betterly

madalu — Fri, 04 Jun 2010 21:16:32 -0000

You've just describe my dream information management tool.

I do think org-mode is the closest I've ever come to such a tool. Research, organization, searching, ad-hoc databases, writing -- all are handled beautifully by org-mode. Last week, I wrote a quick and dirty set of elisp and perl scripts that finds entries in org files similar to the current entry (a la Devon Think). It uses a pretty crude method: tokenize the text of the current entry (with a perl script), remove stopwords (again, a perl script), stem the words (again, a perl script), and use regexp-opt to create a regexp search to be used in conjunction with org-search-view. The results have been surprisingly accurate, though it needs more tweaking.

Re: The Schedule

madalu — Fri, 04 Jun 2010 21:08:09 -0000

I look forward to reading your work, whether on this blog or elsewhere.

Re: Who wants to be a PHP Developer?

madalu — Tue, 11 May 2010 22:23:19 -0000

Alas, I spend so much of my time inside emacs trying to write python and perl scripts to grab data en masse that I forget php even exists. And when I do venture out onto the world wide web, I make sure to browse only static html sites such as your own. :)

Re: A Git of One's Own

madalu — Wed, 05 May 2010 08:10:43 -0000

Sorry to draw this conversation out longer, but here are my solutions:

1. The USB is strongly encrypted, so if I lose it, it's not too much of a worry.
2. I have all the git repos on multiple machines, so if I lose the USB stick or if it fails, I don't lose any crucial data.
3. Re: mount points: I just use a volume label to ensure the same mount point across multiple machines.

Re: A Git of One's Own

madalu — Tue, 27 Apr 2010 21:32:43 -0000

Yeah, the USB thing is a relic of the days when I couldn't count on a reliable internet connection. I was burned one too many times by not having access to my most recent commits --- but I will admit that carrying around a USB stick (even an encrypted one) is risky business.

Re: A Git of One's Own

madalu — Tue, 27 Apr 2010 16:40:22 -0000

Git is great for so many reasons, not least of which is its blazing fast speed. Since it's a distributed version control system, there's *no one right way* to do things. For instance, I keep my computers in sync by pushing to and pulling from bare repositories on a USB stick. A simple bash script is all I need to manage multiple repositories.

A couple of additional thoughts. If you're coming from the svn world (as I did), you'll want to resist the temptation to create a gargantuan repository storing multiple projects. I used to keep all my work in one big svn repo, confident in the knowledge that I could easily check out individual subfolders. With git, there's no easy way to checkout or clone a sub-directory of a repository. It's best (and so much more convenient) to work with smaller, self-contained repositories and to use another tool (e.g., a custom script) to keep them all coordinated.

I solve the binary problem by creating separate repositories for binary stuff I don't plan on changing (e.g., downloaded pdfs). That way, I can still get the benefits of git for keeping my computers in sync. That said, I would not recommend git for managing, say, a video editing workflow.

Re: Objective Whatsis

madalu — Thu, 15 Apr 2010 12:52:11 -0000

Thanks Alaric for that exceptionally clear explanation!

It seems to my coding-challenged brain that OO makes it a lot easier to create "frameworks" that can be easily used/modified by others. It's a way of making code more modular (opening it up to the outside world by making everything an "API"). E.g., the most popular rest web frameworks are OO, which makes them easy to use and extend. IMO, well-written procedural or functional code (especially if it's lisp) can be just as abstract and modular, but the OO paradigm remains the dominant way of thinking about abstraction today.

OO can make simple programs needlessly complex and verbose (e.g., a lot of Java), but it can also really help to clarify large, complex, multi-author programs. All the more reason to choose laguages that offer the best of all the paradigms!

Re: Planned Obsolescence and Gadgets

madalu — Fri, 09 Apr 2010 21:38:55 -0000

My current solution: two ridiculous underpowered computers and no cell phone/MID. ;)

My humble gear (of which I am a bit too proud) = a $200 DIY Atom box at home and a 1 1/2 yr old Atom netbook.

I'm eyeing some of the Android phones on the market, but I'm too much of a keyboard addict to consider a tablet for anything other than media consumption. Alas, my dream machine is more of the micro-computer in a pocket variety. If they could boost the specs on the NanoNote, I'd buy one in a second:

http://sharism.cc/

Re: Content Management Beyond Wordpress

madalu — Thu, 08 Apr 2010 13:10:57 -0000

"We might call this the API-ization or the Unix-ification of web development."

I pray for this fervently every day.

Re: static generation: org-babel is a really neat meta-programming tool that allows you to pipe stuff back and forth between functions written in different languages within an org file and to export the results (text, graphs, data, etc.) to a static html website. So long as you're not creating a social website, this type of static/dynamic hybrid is really nice.

Re: Enterprise Linux Community

madalu — Thu, 01 Apr 2010 14:13:50 -0000

Thanks for these thoughts. I would add that the word "community" is a marketing buzzword, a little bit like "organic foods" or "the cloud". A lot of companies want the good PR and fuzzy feelings associated with the phrase "open source community."

What does one get from contributing to an open source "community" (enterprise or not)? I would say empowerment. Open source offers a DIY education -- whether it's coding, sysadmin stuff, or general knowledge about how computers and networks work. Indeed, by contributing to an open source community, you gain the expertise to use free tools freely and effectively -- and there seems to be a nice little market for that skill. If you contribute to the Fedora project, for instance, you gain invaluable skills about the guts of Linux that you can transfer to other distros/careers/coding projects, though you may never work for Red Hat.

I think the success of the "open" model is made clear by the open source programming languages. By learning a language like Python or Perl you are not only gaining a skill; you are also contributing to the spread and development of the language. Similarly, by learning Linux (whether enterprise or no), you promote Linux everywhere. And, arguably, some of the big, profitable companies that rely on Linux servers have done as much to promote open source as anyone.

Re: Common Lisp, Using ASDF Install With SBCL

madalu — Fri, 26 Mar 2010 10:23:42 -0000

Thanks for these instructions. I'm planning to install SBCL on archlinux soon. I'm still trying to get my head around perl and python first. But those of us with coder-envy and humanities degrees have to take baby steps, right?

Now that you have some experience with common lisp, do you have a sense of how familiar it would seem to your casual emacs-lisp hacker? I've built up some respectable elisp skills and find the syntax very comfortable. How big a leap is it from there to common lisp?

Re: Why Cloud Culture Isn't

madalu — Thu, 25 Mar 2010 00:14:48 -0000

Thanks for the link. The article linked to in the Wired post seems to me yet another example of tech commentary that couples the word "we" with excessively stark dichotomies to prophesy in most portentious tones. "In the past *we* kept our data on hard drives. In the future *we* will keep our data in the cloud!" "In the past we used the outmoded technology of PC and keyboard. In the future we will use tablets for everything!" While the social and technological are of cource intricately interrelated, many commentators simply conflate the two. As you so brilliantly point out, the technology behind the so-called cloud is not all that new (except, of course, for the spread of increasingly robust mobile device). What changes in the way in which we use that technology---i.e., the social and economic dynamics that result from an ever-growing network of connected computers. Granted, mobile devices allow people to connect to the network more frequently and in new ways. But the way some commentators talk, you'd think that the cloud (with the assistance of the mystical iPad) will vaporize the very CPUs on which it runs.

Re: Managing Management Costs

madalu — Wed, 17 Feb 2010 22:06:55 -0000

Interesting post. I would argue that decent indexing and search tools are more important than directory organization -- but that a sane directory organization *combined with* good indexing is even better. Apart from GTD style contexts, I myself rarely use tags -- I find them more trouble than they're worth, since it's very difficult to create a consistent tagging scheme. In fact, I find it much less stressful to place a document related to taxes in a "~/taxes/2009" directory than to dump it who knows where and to rely solely on tagging and/or indexing to find it later. There's something in the human mind that wants things in clear, consistent places. :)

I've tried the "dump it all into one directory" approach before --- and while it certainly makes grepping a bit easier, it also makes it a bit more difficult to keep current projects in decent order. Today I rely primarily on the following tools for organization:

1) recoll - a great Linux search engine (akin to Spotlight)
2) notmuch - a fantastic, gmail-like email search and tagging interface
3) org-mode - need I say more?
4) git - forces me to keep a log and history of my work
5) a directory hierarchy organized around workflow (I have a ~/files directory for current projects. After a project is complete, I use a simple function in emacs dired to archive it with timestamp to my ~/archives directory. E.g., "~/files/the-great-american-novel/" becomes "~/archive/2010-02-17-the-great-american-novel/" when complete.)

Re: Decreasing Emacs Start Times

madalu — Thu, 04 Feb 2010 10:31:46 -0000

I used to split up my files but found that all sorts of cruft crept into the untended ones. So now I use a modular approach within a single file. E.g.,

(setq *flyspell-enabled* t)
(setq *gnus-enabled* t)

(when *flyspell-enabled*
;; .. settings here
) ; end flyspell settings

And so on...

Good point about emacs server. It's funny. I suppose I use the server instead of the daemon because I want emacs to shut down when I close the last frame. The reason: I rely on git and a usb stick to keep my files/mail etc. in sync on different machines. I also rely on emacs timers to fetch mail, rss feeds, and whatnot automatically (but only when I'm working on a particular machine -- i.e., not as a cron job). Thus, I want all emacs initiated activity to cease when I'm done working on a particular machine, so make for smoother synchronization. Alas, the intricacies of customization!

Re: Decreasing Emacs Start Times

madalu — Wed, 03 Feb 2010 09:02:32 -0000

Good stuff. Needless to say, emacs, unlike vim, is meant to be started once.

I myself use emacs server instead of the daemon because I really don't need multiple instances of emacs, just a convenient way to manage new windows (whether in the GUI or console). The new windows pop up immediately. And there are no display issues with emacs server.

I find it convenient to alias "emacs" to "emacsclient -c -a /usr/bin/emacs". This will launch a new instance of emacs if there is not a server running; otherwise it will connect to the server or daemon.

I still probably load way too many packages when first starting up emacs (my .emacs file is well over 3,000 lines long). I save some startup time by automatically byte-compiling that file when killing emacs and testing to see whether .emacs or .emacs.elc is newer when starting emacs. One of the gazillion hidden secrets in Emacs is the wonderful byte-recompile-directory, which will automatically byte-compile .el files in a directory that are newer than their .elc counterparts. I find the following hook particularly convenient:

(add-hook 'kill-emacs-hook
'(lambda ()
(byte-recompile-directory "~/mystuff/elisp/")
(byte-recompile-directory "~/mystuff/emacs/")))

Re: Beyond Lists in Org Mode

madalu — Wed, 20 Jan 2010 14:34:09 -0000

Thanks for this great post! As you might guess, I can't help but chime in whenever emacs is mentioned. :)

I'd agree that the sweet spot for org-mode files can be hard to find. I have a file per "area of interest/responsibility" and tend to reserve level one headlines for projects. (Tags + filetags + agenda filtering = a whole lot of awesome!)

My solution to the problem of messy files was to create an "inbox" file in which all new remember items are deposited. I simply refile the items to the appropriate file and project when I process the inbox. I find the following settings particularly helpful:

(setq org-refile-allow-creating-parent-nodes 'confirm)
(setq org-outline-path-complete-in-steps t)
(setq org-refile-use-outline-path 'file)

Funny about archiving.... I use org-archive-subtree all the time. I find it quite convenient to clean out my org-files and to keep the agenda commands relatively speedy. I have my archive files automatically divided by year, which keeps things nicely organized (especially since I have a very high volume of notes):

(setq org-archive-location (concat (format-time-string ".%y") ".%s_archive::"))

I've stopped writing content in org-mode. Instead, I've begun to use my project subtrees as the organizational hub for all my writing. Org-attach allows me to jump quickly to the relevant documents.

But the crazy thing is that despite all this goodness, I still occasionally am tempted by the talk of Arch minimalists to go back to vim. ;)

Re: End User RSS

madalu — Mon, 11 Jan 2010 23:13:16 -0000

Forgive me for my predictable emacs boosterism:

I use gnus in emacs (no surprise there!) with shimbun.el (part of emacs-w3m). This helps to solve some of the problems you mention in the post, though it creates new headaches (e.g., some bugs, fussing around with elisp, weird characters, etc.).

Shimbun is basically a web-scraping toolkit. If a feed you read doesn't include the full text, then you can instruct shimbun to pull in all the text from the site (minus the cruft). The shimbun rss-blogs simplifies this quite a bit. But you can also set up web-scraping for websites without feeds, etc. Or you can instruct a shimbun to combine a multi-page newspiece/post into a single article in gnus. This can require some heavy-duty emacs lisp. Thankfully, there are lots of pre-defined shimbun (NYT, slashdot, BBC, etc.). Indeed, shimbun is ideal for newspapers that have truncated rss content and multipage web articles.

Gnus' outstanding scoring capabilities can be set up to prioritize the most read groups.

Needless to say, receiving the full text of articles in emacs has radically simplified my daily newsreading. If I want to remember and article, I clip it into org-mode (which I have set up to save the permanent url).

Since I also read tech-related mailing-lists via gmane's nntp server (which saves me from having to collect and filter thousands of emails myself), I can read all sorts of news in the same place.

It's astonishing that there aren't user-friendly tools like this. But then, the more one uses emacs, the more one realizes that the rest of the computing world is still catching up. ;)