# C[omp]ute

Welcome to my blog, which was once a mailing list of the same name and is still generated by mail. Please reply via the "comment" links.

Always interested in offers/projects/new ideas. Eclectic experience in fields like: numerical computing; Python web; Java enterprise; functional languages; GPGPU; SQL databases; etc. Based in Santiago, Chile; telecommute worldwide. CV; email.

© 2006-2013 Andrew Cooke (site) / post authors (content).

## Leaving GMail...

From: andrew cooke <andrew@...>

Date: Sun, 7 Mar 2010 15:42:24 -0300

I am increasingly concerned about Google constructing a detailed model
of "me".  While I have nothing against them as a company, I am not
that keen on anyone compiling a large pool of information about me -
particularly one over which I have little control.

A month or so ago, I tried switching to Bing for searches.
Unfortunately, Bing is significantly worse than Google and good
quality search results are critical (I am a software engineer with a
poor memory - search solves almost every problem I have).  So
switching search away from Google search failed.

The next alternative, then, was to change webmail providers.  After
checking out the summary on Wikipedia, I decided that Runbox was best.
But when I opened a free, trial account I wasn't convinced (it's
quite solid, but I had some worries: the folder display would not
automatically update to reflect incoming messages; it wasn't clear how
to apply filtering to existing (as opposed to newly arriving) emails;
no tags).

After that, I started considering how to create a solution myself.  I
ran Squirrelmail locally for many years, and I now have a server that
could host it (in fact, my server provider also has webmail) - but
Runbox was better than that.  And admin work isn't appealing.

So I paused for a think.  A think about GMail's tagging of messages,
and how you could do something similar with maildir and soft links on
mairix.  Mairix is an old-school command line program that lets you
dynamically construct maildirs (folders of email) according to a
search.  And it's fast (it pre-constructs an index).

Some background on hardware:  Currently at home the only computer in
the house that's on 24 hours a day is a little, low-power box that
serves music, works as a firewall, etc (it has a Via C7-D chip - think
early netbook).  It contains all my email in a bunch of directories,
sorted by date (these are pulled from GMail every 3 minutes) - 600,000
emails going back years.  I can connect to this box from anywhere over
ssh (I use a dynamic name service to locate it).

So I started wondering about dropping webmail and going with mutt in a
terminal window.  I could use mairix to generate folders like GMail's
tags, and also to do searches...

...and the good news is that mairix compiles and appears to work just
fine.  Even on the under-powered box, with over 1/2 a million mail
messages, it's instant.  With a little scripting I can create a bunch
of directories that are equivalent to my GMail tags, and I can also
search by hand.

Mairix (from git) installed and bit with no problems.  And indexed
everything in about 10 minutes.  I've been a little frustrated with
GMail's search recently; my first impression of mairix is more
positive - it really works!  The main remaining worry is how to handle
attachments - with webmail that's trivial; it's not so much fun via a
remote tty.

Anyway, that's the latest news.  I'll report back once I have some
serious experience.

Mairix - http://www.rpcurnow.force9.co.uk/mairix/

Andrew

### Mutt Working Well

From: andrew cooke <andrew@...>

Date: Tue, 9 Mar 2010 02:12:27 -0300

This is working out well so far.  The hardest part was getting mail delivery
working with my server host - MTA's send to MX addresses, but their SMTP relay
needed to be addressed directy (the solution is to use [] in postfix's
config!).

Anyway, I thought I'd describe how I'm using Mutt so that it duplicates the

When mail arrives it is stored by month.  My default folder is the current
month, which is similar to "inbox".  Apart from spam (which is filtered with
spamassasin), everything is stored in this way.  Messages are threaded and
sorted by date, which gives results similar to the way GMail groups related
messages.

In addition, I have some additional folders that are created by mairix.
Mairix runs in the background every 20 minutes to update its index, and every
time mail is delivered to update the extra folders.

There are two folders.  One contains work-related items for the last month -
this is created by searching for the host names of clients.  The other
contains friends and family for the last month, created by searching for email
the equivalent of GMail's tags.

In the work folder, my "from" address is automatically changed to reflect my
work email.  In other folders it reverts back to andrew@...  GMail has
no equivalent for this.

If I need anything else, I can invoke mairix from within mutt, and the results
appear in a third folder.  This is the equivalent of GMail's search (except it
appears to be better - I found GMail search strangely unreliable, but I could
never pin down exactly why).

Now we'll see if posting here still works...

Andrew

### Spam Filtering Details

From: andrew cooke <andrew@...>

Date: Fri, 12 Mar 2010 11:13:30 -0300

See this post for more on filtering spam -
http://www.acooke.org/cute/EfficientS0.html

Andrew