[[!img notmuch-logo.png alt="Notmuch logo" class="left"]]
#Using notmuch remotely#

##Why?##
It is hard to keep nomuch tags in sync across multiple instances of
notmuch, on multiple computers. Though you can do this with "notmuch
dump" and "notmuch restore", it is often preferable to be able to use
notmuch on a remote computer as if it were present on a local
computer.

The following guidelines show how I have accomplished this. It isn't
perfect, but it works pretty well, and allows me to access notmuch on
my home computer, using only an emacs client on my netbook or work
computer, a trivial shell script, a few settings in my .emacs, and
ssh.

Note that this is all something of a hack, and future versions of
notmuch will likely make all of these steps much more
transparent. I'll note particularly which things should become
unneccessary with future version. At the moment though, this does
work, and might enable some of you to use notmuch away from your
primary computer.

##What you will need##
You will need to have the following items in place:

1.  a working notmuch on one computer (let's call that computer
"server"). The notmuch should be at least version 0.2.
2.  a working notmuch emacs interface on another computer (let's call
that computer "client")
3.   password-free login (public key authentication) from client to
server. [Here](http://sial.org/howto/openssh/publickey-auth/) is a
good page on how to set it up.
4.   a reasonably fast connection. (This isn't really *neccessary*, but
if your connection is too slow, this won't be very pleasant to use,
and certainly won't seem transparent.)

##Write a simple wrapper shell script##
Now we will need to write a simple shell script that replaces the
call to the notmuch binary with a call to notmuch over ssh. 

Note that this shell script also pauses briefly after every ten search
entries. This is currently necessary so that the emacs process-filter
doesn't chop off messages. It's an obvious hack, and hopefully won't
be necessary in the furture.

	#!/usr/bin/env bash
	
	SSH_BIN="/path/to/ssh/on/client"
	USER="user_name"
	HOST="server_name"
	NOTMUCH_REMOTE_BIN="/path/to/notmuch/on/server"
	
	if [ $1 = "search" ]; then
	    COUNT=0;
	    OUT=`$SSH_BIN $USER@$HOST $NOTMUCH_REMOTE_BIN $@`
	    echo "$OUT" | while read line; do
		COUNT=`expr $COUNT + 1`
		echo "$line";
		if [ $COUNT = 10 ]; then
		    sleep 0.1;
		fi
	    done
	else
	    $SSH_BIN $USER@$HOST $NOTMUCH_REMOTE_BIN $@
	fi
	
Save this to a file, "remote-notmuch.sh", in your path.

##Configure your emacs client##

Add the following to your .emacs (this is on your client machine):

    (setq notmuch-command "/path/to/your/remote-notmuch.sh")

At least until 0.3 or 0.4, you will also need to add the following. It
should become unnecessary pretty soon though:

    (setq notmuch-remote-host "user_name@server_name")
    
    (defadvice notmuch-show-get-filename (around
    				      notmuch-show-get-remote-filename
    				      activate)
      (setq ad-return-value (concat "/ssh:"
    				notmuch-remote-host
    				":"
    				ad-do-it)))

The purpose of these lines is to allow emacs to have access to the raw
files, via TRAMP, so that it can extract attachments and parse
HTML. Work is afoot to make notmuch handle these tasks itself, so this
part should soon be unecessary.

##Problems##
Some things won't work perfectly, and there might be some unexpected
mismatches between normal usage and this sort of usage. If you're
using this approach and run into any problems, please feel free to
list them here. And, of course, if you improve on any of these
approaches, please do edit this page and let people know!