ssh server cygwin setup

So you want to be able to ssh into your windows box? Don’t know your windows password, but have admin rights? Here’s how: use the “net” command to create a user with a password. Then install openssl-server from cygwin setup, then sync the users

mkpasswd -cl >/etc/passwd
mkgroup --local > /etc/group

Then fire up the server net start sshd and it should work. See for more tips (most of which may not be necessary).

If you want to be able to actually edit files, etc. then you are going to want to login as your ‘normal’ login name.  So <code>passwd normalusername</code> and enter a new password.  Now you don’t need to know the windows password 🙂 .  Next if you want to not have to type in a password, create a key pair which allows you to authenticate with it.  Download putty, run puttygen, click generate, move the mouse to create a new key pair, click “save private key” (give it a passphrase–this protects you if a bad guy gets it), next copy the text from the topmost window (in grey).  Now login to localhost <code> ssh user@localhost</code>, create a ‘.ssh/authorized_keys’ file <code>mkdir .ssh; cd .ssh;</code> and put that stuff in it (here’s one way <code> echo “paste it in here” >> authorized_keys</code>.  Now explore to where you put your private key, double click on it (fires up the ‘key manager’), and now logging in to username@localhost should be automatic when you supply the username.  (Putty can apply the username automatically, if desired, see options).  Now to get it really automatic then put a link to that private key in your startup folder, and then never worry about it again 🙂 Good luck!

Now you can make it so

apache stdout

So…want to know where apache writes text which is written to stdout? (i.e. “print “hello””).  It (I think) outputs it to the console window of the process that started apache (i.e. init, if auto-started), so that stuff is toast!  Text written to stderr will however go into the file “error.log” in the logs directory, so that’s a good way to still see your output. GL!

openid RoR woes

So…you download the RoR openid plugin, install it, and…

nothing seems to work?  The reason may be that you need edge rails installed (apparently).

The error I would run into is that open_id? method was not defined, or what not (etc., etc.)

So go and get it instead from where there is a baked sample. Otherwise you might get stuck after gem install ruby-openid --include-dependencies
ruby script\plugin install open_id_authentication
. GL!

php mail

So…you want to send php using the mail function, you load it up and…nothing?  First check if sendmail works mailx -s “subject” (type, then Ctrl+d, then enter, to end).  If that works then you may be getting the following error:

“Recipient names must be specified”

This actually originates from sendmail.  What it meant for me is that you *must* include a “from header” for mail, at least on Oracle Linux.  Oh BTW installing php4 on oracle linux seems impossible.  You can edit the config file all you want to try and include those /usr/lib64 includes, then it ends up saying “-lxml” on the command line and then doesn’t compile. Nuts. 🙂

Using mysql in oracle Linux

Turns out that mysql IS by default installed in oracle Linux. How to set it up:

First let’s turn it on
go to /etc/rc.d/rc5.d and copy K36mysqld to S36mysqld [i.e. set it to start it up, too]. then run it sudo s36mysqld start Mysql is officially running.
Next setup root (stolen shamelessly from
run mysql -u root
grant all on *.* to root@localhost identified by 'insert_pass_here' with grant option;

(password above is your root pw)
and test the login again with the password

./mysql -u root -p

and enter your password.

opt and blank php pages

So…if you go to a php page and it’s blank…do not fear!  This may not mean php is broken.  Go to the top of that php page and put “hello” or what not to see if it actually php or apache that is broken, or if php is coughing on something you are running later (and returning nothing). GL!

RoR scaling

So most people complain about Rails not scaling (well, rather, fear it).  It seems that with 10 servers you can host about 45 million hits a day.  And hardware load balancers.  So…that’s kind of expensive, but hey 🙂  lists the post.  Also at they list how they scale using mongrel and “only 5 machines” — ok so it scales, but with tons of machines?  How well do other protocols scale like that?  I would like to see a list of ‘scaling’ on each protocol, with the best of practice.  That would be cool.  How about a ‘scaling cook-off’?

Roger's meanderings, notes to himself, bug reports, and other things