Apple OSX Mail Program Changes the Port on Its Own

I use Mandrill for my outbound SMTP. For SSL, it operates through port 587.

I have had many occasions where my connection fails for no apparent reason. I'm just living my life and sending email and then, poof!, an email tells me it can't send with that server. I go through debugging and re-enter the credentials and it would work.

I started thinking "last time it happened, wasn't it the port number that was different then, too?" So, I changed the name of the server to "Mandrill, s/b port 587" so I would know.

It happened again today. I went to look at it the settings and there was a change to port 486. WTF?


I changed it back to 587 and it worked.

I have changed the setting "Automatically detect and maintain account settings" to be off. I'm going to be that stops the problem. If not, I'll update here. If it's been a long time since 5/15/15, you are safe in seeing that as a solution.

bbedit as commit message editor for Git

I had this

export EDITOR="bbedit -w"

and, as expected, Git opened bbedit.

But, it also said

Aborting commit due to empty commit message.

Screwy.

I found this:

git config --global core.editor "bbedit --wait --resume"

and put it here so it would be remembered.


Dave Winer Strikes a Blow for Freedom

I started my web career with this guy's software. He's an original blogger and podcaster whom I've been reading since the nineties. He's done a lot of interesting things but now he's doing something important.

He has put together a Javscript/Browser blog editing environment that is not tied to any predefined external service. He's making a big point of keeping all open source. One of his major ideas is creating ecosystems. I think this will establish an important niche.

The reason it's important is that it will allow us to establish a web presence without relying on some big company's software. For the most part, anyone who wants to blog or otherwise express him or herself on the web has to use Facebook, Twitter, Blogger, hosted Wordpress, Tumblr, etc. That is, you put yourself at the mercy of their terms and conditions. Worse, your information is stored on their servers where it is probably hard to retrieve for posterity, reuse or backup - if you can get it at all.

Dave's thing is intended to counter that. As I understand it, pre-release, he has defined a protocol for storing data from the browser that can be implemented on any server you want. He also has created a NodeJS component that implements the protocol. All open source.

With his other Javascript things (he calls them "snacks"), he has made configuring and using them an entirely browser-based experience. It sounds like this might require creating a storage location this time. We will see.

However, he has often expressed a philosophical viewpoint with which I agree heartily that being free requires that we be able to communicate on our own terms with technology that is sufficiently decentralized that we cannot be shut down. I'm looking forward to his release of his MyWord Editor Project.

http://tqwhite.org/?89E805

Dave provides additional explanation.

http://tqwhite.org?41A5AE

MSSQL Service Broker Does Not Work

It's time for a project I've not seen in awhile to get some love. It's .NET MVC running against SQL Server.

I fire it up. After some really annoying compiler reference issues (how does this happen on a project that has not been touched?), I get a dev version up and running.

I want to work on a copy of the database but I also want to leave the original up for reference as I move forward. So, I back up, make a new database and restore into it. I run my initialization script that, mainly, establishes a service broker. Edit the connection string to point at the new database, and, drum roll. . . . .

No love. The service broker is not configured properly.

This stuff worked just fine last time I used it and for years before that. I check a billion things. I recreate the database making sure I didn't screw anything up. I turn on the debugger and look at it from the program's viewpoint. Nothing does any good.

I am beat. After another round of dejected facebook or twitter or something, I am just staring at Management Studio. I, once again, look at the properties of the database. Goddam, Service Broker is turned on. I look at the original database again, Service Broker is turned on.

But I notice the GUID for the first time.

Bottom line, the Service Broker GUID was restored along with the rest of the database. For no particular reason (probably a guy who worked here years ago), I was setting ENABLE_BROKER.

A little google later and I learned about NEW_BROKER. Done. Works.

If you have a service broker that doesn't work, won't respond or otherwise gives you a big pain, check the database property/options and scroll down to service broker. If the GUID is the same as in another database, there's your problem.

OR, you could immediately try, ...

ALTER DATABASE [database name] SET NEW_BROKER WITH ROLLBACK IMMEDIATE;

And go to lunch.

Secure Email - OSX and IOS

(I never went back and edited this. That means it's probably a little rough and may be confusing.)

Secure email requires encryption. Encryption requires two keys. One is called a public key, a sender can use this to encrypt a message. The other is called a private key. It can be used to decrypt a message that was encrypted with the corresponding public key.

When you send secure email, there are two functions. One is called 'signing'. In a signed message, the mail program uses the recipient's public key to create something called a digest. The recipient can open the digest (with their private key) and verify that the email is exactly the one that was sent. The other security function is called encrypting. In this case, the sender uses a public key to put an encrypted form of the message into the email and send that instead. The recipient uses his or her private key to decrypt the message into a readable form that security people call plaintext.

When you send an encrypted or signed email to a list of people, a different key is used for each person. A different encrypted package is sent to each. Each uses his or her personal private key to open the message. You cannot send an encrypted message to a list of people that includes someone for whom you do not have a public key.

Managing the keys is done with something called certificates. The certificate contains both the public and the private key and information that allows the mail program (in cases where its important enough to pay the certificate company) to verify that you are you and that this is your key.

To start, get a free certificate at:

https://www.comodo.com/home/email-security/free-email-certificate.php

After executing the signup, it will email your the certificate.
- Double click on the attachment in the email from Comodo tol open Keychain Access.

It will automatically install the certificate.

To verify that it all works, quit Mail.app and restart it. Create a new email and you will see new icons on the left of the subject line.



That blue check means your emails are now signed. If some hacker messes with the message before it gets to me, I will get a notification that the message is not the one you sent. The grayed out lock symbol says that you are not yet able to send an encrypted email, yet.

To be able to send encrypted email, your email program has to have the other person’s public key. Mail.app collects these when it gets a signed email. To send an encrypted email, ask the person (I have one so I'm a good option) to send you a signed one and reply (or Mail.app may already have it). The lock should turn blue in a new message to that person.

But, when you receive an encrypted message (eg, when I reply to your first encrypted message), you will see that you cannot read it on your iPhone. The problem is that your iPhone needs to have your certificate as well.


NOTE: YOU DO NOT NEED TO DO THIS. THE EXPORT FROM KEYCHAIN CAN BE SENT. I DON'T REMEMBER EXACTLY WHAT I DID, SO I AM LEAVING THIS PART IN.

To update your iPhone with a certificate, you need a new program called the iPhone Configuration Utility.

Download it from  

http://support.apple.com/kb/DL1465

Install it like any other program. It gets installed into "Applications/Utility/IPhone Configuration Utility".

Open Keychain Access
- Find the certificate that was installed before in the ‘login’ keychain. It is named as your email address. If you have old ones, it will probably be the one with the latest expiration date.
- Select it and choose Export Items from the File menu.

You will see a file dialog. In addition to files, it will show a popup men under the file list.
- Choose “Personal Information Exchange (p12)”
- Save the file someplace easy to find again (Desktop is good. You can throw it away after this is done.)

Open iPhone Configuration Utility.
- Select “Configuration Profiles” from the palette in the left column
- Click the New icon on the top left.

This will show a new set of controls in the lower left window.

Select the General item.
-` Make up an informative name .
- Then make up a unique Identifier. For this purpose, I suggest something like com.domain.emailCert.yourEmailName.

Leave Description and Security as they are (blank, or whatever you want to type, and Always).

Next, scroll down from General and find Credentials.
- Click on the Configure button that shows and you will see a file dialog. Choose the p12 file you made before and Open.

The final act in Configuration Utility is to click Share, next to the New icon.
- Select None from the Security popup and Share…. This will take you to Mail.app.

There you will see a new mail message with a file attached. Email this to yourself.

The last step is to open the email on your iPhone and tap the certificate.

It will open.
- Click the button, top left, that says Install. It will ask for your phone’s password.
- Enter your password and click Done. It will tell you that the profile is not signed.
- Click Install. It will ask you to do it again to confirm.
- Click Install again (this time red and at the bottom).

It will return you to your email message and you are done.

(You can look at it or delete it by going to Settings->General->Profile. It’s way down at the bottom).

Now you can find the email that you couldn’t read before and you will be able to read it.


SSL Certificate for Apache

I just renewed the certificate for a client web store. This is a task that drives me crazy because I do it so infrequently that I can never remember the details. It's made worse by the fact that I use InstantSSL, a company whose documentation and practices are incredibly confusing. Also, cheap.

Here's what happened.

I followed the usual instructions for generating a Certificate Signing Request. This generated two files, a .csr and a .key (for later).
I used a command line like this:

openssl req -new -newkey rsa:2048 -nodes -keyout myDomainName.key -out myDomainName.csr

It is entirely unclear to me if the name of the files makes any difference. Also, in the dialog that openssl conducts to construct the CSR, it asks for "your name". It means 'your domain name'. May be obvious but I thought that maybe it was picking up the domain name from the command line.

The purchase process for InstantSSL is pretty easy. They are really good at taking money.

Since I screwed up the domain name thing, I had to interact with tech support and give them a new CSR. This was annoying since it just said I needed to do it, not how. Eventually I had looked around long enough that I felt like I had to call. Turns out you have to construct a support account and treat send it through the support website along with your order number.

The next step was validation. It does two things. It looks up the company phone number (in this case using Dunn & Bradstreet). I had to work with the client to be there to answer the phone and gather the 'validation code'. Entering this into the phone call tool triggers another email (this came to me, as the email address on the purchase) with another, longer validation code. This had to be put into the order control panel.

While there, I saw that it still wanted to send email to admin@, an address that does not exist, for additional validation. Turns out that, in the order control panel, you can change this to one of several other options. I don't know where they come from but, one of them was me. That sent another validation code to put into the order control panel.

And then I got an email with a .zip full of certification. Expanding this got me two files, a .crt and a .ca-bundle. I also needed the .key file made along with the CSR.

These were put into folders specified in the Apache config file. In my case, this was a virtual host file for that site.

There are three files and they need three entries for Apache configuration:

    SSLCertificateFile /etc/ssl/certs/sales/myDomain.crt
    SSLCertificateChainFile /etc/ssl/certs/sales/myDomain.ca-bundle


    SSLCertificateKeyFile /etc/ssl/private/myDomain.key

On my system, these are root files with permission 644. I do not know if this is ideal. There might be some more secure arrangement. (That's what comments are for.)

After a server restart, the browser security violation went away and things look pretty good.

Now you know.





Debugging a Web Type Font Served from a Different Domain

A client came up with a type face that she found on some site. The supplied file type was .ttf (truetype).

Because I treat my public_html folder as source code, I always keep a third level domain for static file server, eg, static.someDomain.com, to hold files that need to be directly served. I expect that this might have worked more easily if I had been able to/chosen to serve the type file from the site's public_html folder instead of the different domain.

I put the .ttf on the static domain and added this coding to the web page:

@font-face {
  font-family: 'demoFont';
  src: local('demoFont'), url(http://static.someDomain.com/demoFont.ttf) format('ttf');
}

Even though the internet said that Firefox wants .ttf , it didn't work.

I have used Google fonts before so I checked that out for comparison.
@font-face {
  font-family: 'demoFont';
  font-style: normal;
  font-weight: 400;
  src: local('demoFont'), url(http://static.someDomain.com/demoFont.woff) format('woff');
}

I find a site that does ttf-woff conversion ([1] below). It works great but, after I copy it in and edit to use the google version of @font-face. No dice.

Reading the internet ([2] below), I got the idea to look at the headers ('curl -I domain.com') coming back from google and my static domain specifically to see about 'Access-Control-Allow-Origin'. I found that google's header included "Access-Control-Allow-Origin: *" and mine did not.

The website suggested using .htaccess to add the goodies to the header:

<FilesMatch "\.(ttf|otf|eot|woff)$">
    <IfModule mod_headers.c>
        Header set Access-Control-Allow-Origin "*"
    </IfModule>
</FilesMatch>

(Actually, I added the '|woff')

Checking the header... No dice. The line does not appear.

I notice mod_headers.c and check my Apache config. Not there. Navigate to /etc/apache2/sites-enabled:

ln -s ../mods-available/headers.load headers.load

service apache2 restart

Bada Bing. Bada Boom!

The header is there and the the typeface shows. Win! And, for good measure, it works in Chrome, Safari and late model Internet Explorer.


[1] - http://everythingfonts.com/ttf-to-woff
[2] - http://stackoverflow.com/questions/2856502/css-font-face-not-working-with-firefox-but-working-with-chrome-and-ie










SSH fails to use RSA valid RSA key because the account is not owned by correct user

For various reasons, there are three accounts on this Linux server that I want to access with SSH using my RSA key pair. I generate them and copy my key into each account, tune up the authorized keys file and prepare for satisfaction. What I got was much less. Only one of the accounts logged in using the RSA key. The other two ignored the key and insisted on a password.

The first thing I did was install the key on a couple of other servers and demonstrated that the keys were good.

Then I had my SSH client show me the transcript and did one million things. I now know a ton about the sshd file. I changed permissions, did careful reinstallations, I spent hours to no avail. I consulted my genius son. At the end, I was still getting this:

debug1: Next authentication method: publickey
debug1: Offering RSA public key: /Users/tqwhite/.ssh/id_rsa
debug1: Authentications that can continue: publickey,password
debug1: Next authentication method: password

That's right, it was actually trying to use the key and it was being rejected. Then, it would ask for a password. I was stuck and gave up. Actually, I ran out of time.

Some time later, I was building a new Linux server at my day job and, damned if I don't run into the same problem. So, I dig in and try again.

This time I'm in a slightly different context and, I'm not sure how I got there but found out about this:

sshd -d

It turns on debugging on the server side. (It's a little bit annoying because, after one try, it breaks something and I had to reboot to get everybody cleaned up and working again for another try. I'm sure there's a better way but I don't really care. Rebooting was ok on this server.)

To use it, I logged in with one terminal window, turned on debugging, then tried to log in with another window. The debug window showed a transcript of the server side of the experience.

I didn't grab a complete transcript but it did include this phrase:

bad ownership or modes for directory

So, I cycle back and make sure that my .ssh directory is set correctly:

chmod go-w ~/
chmod 700 ~/.ssh
chmod 600 ~/.ssh/authorized_keys

Still no dice.

I'm thrashing about, feeling desperate and miserable and, suddenly, Voila!, I notice that the freaking account (root, in this case) is not owned by its correct user. I know, wtf?, but there it is. For some reason, the image used to create the server has weirdness on the ownership of account directories. I reflect on "bad ownership or modes" and, moments later,

sudo chown root:root /root

and, bada bing, it works. I go to the server that was killing me previously and, who knows why, the accounts that fail are not owned properly. More chown and life is good.

Awesome Experience with DigitalOcean

[UPDATE2]

DigitalOcean solved the problem and made life good again. We achieved understanding of the specifics of my need and that it is a valid one. Their preferred solution will require changes to their system that might be made in the future. In the meantime, they have agreed that I can have more than one account. That means that I can use DigitalOcean in the way that suits my needs.

Almost more importantly, it means that I can love DigitalOcean again. I told the guy that I feel silly because I actually had hurt feelings and was sad at the prospect of not feeling the DigitalOcean love. I guess I feel even sillier at my elation at having them not only satisfy my need but actually exert the effort to go through conflict to a happy resolution.

The title of this post has been changed from "Rough Experience" to "Awesome Experience with DigitalOcean".

[/UPDATE2]


[UPDATE]

I heard from a different Digital Ocean guy this morning. He was incredibly nice, clearly had read and paid attention to what was said last night and made suggestions for ways to solve my problem. I don't know what the resolution will be but, I'm leaving this post up both so he can read it to understand what I'm trying to do and so that the rest of the world can know that these guys are not only trying to provide an awesome service but are willing to reverse course when they see that they are injuring a customer relationship. I think that might even be more important than getting it right the first time.

[/UPDATE]

Digital Ocean kicked me out last night when I tried to start a new client account using my own credit card. They told me that they are "not comfortable" with me using my credit card for more than two accounts (I have another client I put there already). Huh? Not comfortable using my 100% valid credit card to pay for multiple accounts?

My Situation

It turns out that there are lots of people who want a minimal website and have no way of getting it. When a friend or family member knows one of those people, I sometimes end up making a small website for them. For years, I've just hosted the site on one of my servers. They never have any real traffic so it doesn't cost me anything.

But, I hate it. I have some sites that have been in my life for years and, if I want to change servers, I know I have to move those sites, too. What I really want is to be able to quickly fire up the site and then hand it off to the person. Most of them are happy to pay a few bucks to keep it alive. (I don't take the money because I don't want to do the paperwork.) I'm happy to help again if they need it, but I really would like to have it be their responsibility.

I've thought about just pushing them to a shared host, but, I've not seen one that didn't require a whole other learning curve. More importantly, I manage my software with git and no command line is a problem. And, I put parameters in my virtual host file to drive my software. Putting these sites a virtual host that would support how I want to work is too expensive for them.

Enter Digital Ocean.

These guys offer a very straightforward system to generate a virtual private server on the web. Their smallest instance comes out to about $5.00 a month. I can have a website based on my software up and running there in about twenty minutes.

Aha! A solution to my problem. I can quickly generate the new, tiny site. Because it's cheap, I can afford to host it as a gift for awhile or I can pass the account on to the person or I can invoice them and manage it myself. In all cases, if the client wants to use some other web guy, I can just hand over the password. Or, if I don't want to mess with it any more, I can just hand over the password. Or, as is the case with one more substantial client, I can fire them – and just send them the password.

The point is that I can, in an email, give the site to the client and be done with it. I do not have to do a whole project to change their DNS, initialize a server, blah blah blah. Send the password. Be done.

Opportunity Arrives!!

This notion came to me this summer. Having my own Digital Ocean account, I fired up a new account for a new client and started to work. This one is a little larger than the main use case but, the principal of being able to turn the whole thing over to them when appropriate still applies. Then a friend asked me if I could help out one of her friends.

Sure!, says I. I have this awesome new service available. I'll fire it up tonight. He can send me sixty bucks for this first year of service and Bingo!, we will have his little site working before sunup.

Disappointment Follows

So, I hit Digital Ocean and create the account. I enter my credit card and, Bam!, the site tells me their fraud detector has found me wanting. The messages are all nice and unthreatening. Almost immediately, I get an email asking why I made more than one account with the same credit card. I think, "oh! I can understand that", and explain my situation.

Imagine my surprise when that's not enough. (No! You can't buy three things from me using the same credit card.)

I engage with some customer service guy. He tells me that they have "standard procedures". He asks how we should resolve this. Thinking, "Really, your standard procedures say, 'don't let TQ give you an extra five bucks a month?"

Instead, I tell him that it makes sense to have his software detect my anomalous behavior but, having investigated me and found me to be a real person, it seems like the 'standard procedure' should be to green light my account and let me get to work.

"We are not comfortable"

But, he ignores what I wrote (It's weird. when I reviewed the emails this morning, it's like I was talking to an automated system. He never actually responded anything I said.) and, after asking me to repeat my earlier explanation (was he really paying that little attention?), tells me that they are "not comfortable moving forward" with more than two accounts.

Where I come from, "not comfortable" is the same as "we don't feel like it." He offers no real reason. In my ire, I assume that he doesn't really care how I feel about, at least not enough to tell me the real reason – or, more likely, there isn't one. It really is that they value their procedures over my business.

I told him that I was bummed out about his discomfort and that he could delete the account I made but won't be able to use. He told me that he wasn't going to do the deleting. I have to do it. He explained the procedure I should go through.

I believe the real meaning of that was, "I just told you we are not going to work with you because we don't feel like it. What makes you think we are going to do anything for you. Now, you go clean up our server."

Be Cautious

I have come to really value companies that want to have a good, supportive relationship with me. I send email to the Bare Bones software (make of bbedit, my code editor) and get a response from the principal guy. My main server host, Server Grove, tries to help me out when I contact them. Their main guy sometimes gets involved. There have been occasions when he actually did stuff that I know he didn't have to do to help me. I feel good about those companies and tell everyone how much I like them.

Digital Ocean has such a nice website and a good service. I thought I had found another company to solve a problem that has been bugging me forever.

Not. Bummer.