Java: execute program without blocking

A long while ago that I had done some Java programming... How to run a program from within Java in a decent manner. While looking around for some sample code, most solutions use the Process.waitFor() method to wait for the process to terminate. But that will usually block forever as process writes data to stdout or stderr and nothing reads that output.

One option is to use a separate thread to read stdout/stderr. I opted for an even simpler approach: temporary files:

execCommand = execCommand + " > " + stdoutFile.getFileName();
execCommand = execCommand + " 2> " + stderrFile.getFileName();
// command/program > stdout-temp 2> sterr-temp
Process p = Runtime.getRuntime().exec(execCommand, null, currDir);
int exitValue = 0;
boolean isRunning = true;
int waitSeconds = 30;
while(isRunning && (waitSeconds > 0)) {
try {
exitValue = p.exitValue();
isRunning = false;
} catch(IllegalThreadStateException e) {
// process is still running, wait 1 second
try {
Thread.sleep(1000);
} catch (InterruptedException e1) {
// ignore
}
}
waitSeconds--;
}
if (isRunning) {
p.destroy();
exitValue = 9999;
}
stdoutFile.delete();
stderrFile.delete();

Amazon pub/sub in the cloud

Amazon keeps extending its cloud offering. They have just added Amazon Simple Notification Service (SNS). SNS is a publish/subscribe mechanism.

Integration-As-A-Service
As explained in earlier posts, I expect Integration-As-A-Service to become more important. One of the larger players (Amazon, Google, EMC, Cisco, Microsoft, ...) may one day come up with a wonderful solution for Business-2-Business communication between organizations.

When I first learned about Simple Queuing Service of Amazon back in 2006, I intially thought that SQS could serve as a transport mechanism for B2B communication. But that didn't work out. As the message size of SQS was very limited, data first had to be stored on S3. Authentication and authorization were also very limited.

So I looked around in the SNS documentation to see what SNS actually is and see if it can serve as a basis for B2B communication. Amazon thinks SNS is usable for B2B or application integration:

Application integration: Amazon SNS can be used in workflow systems to relay events among distributed computer applications, move data between data stores, or update records in business systems. For example, in an order processing application, notification messages may be sent whenever a transaction occurs; a customer places an order, the transaction is forwarded to a payment processor for approval, and an order confirmation message is published to an Amazon SNS topic.

Some facts

• Messages can be published over HTTP, HTTPS, E-mail or SQS
• Proprietary solution/mechanism, not based on any standard (no AS1, AS2, SFTP, WS-Notification, WS-Eventing, ...)
  
• Messages are (again) limited to 8KB. Just like SQS: too small.
• Authentication is based on AWS accounts, so also every subscriber requires an AWS account, hindering factor.
• Messages are pushed, not polled. This is good for performance. For polling, use SQS.
  
• But when pushing, the subscriber must expose a web service or mail account. How to secure this: no authentication from Amazon to endpoint receiving notifications; no basic auth, no support for client certs, ...
  
• Messages are signed by Amazon. This is good, very good. Signing is based on HmacSHA256.
  

Conclusion:
Nice and interesting, but not good enough... In particular the message size remains a blocking factor.

Questions left:

• What happens if messages cannot be delivered for a longer periode of time? E.g. when a subscriber disappears?
• How does a message that is published over HTTP exactly look like (signed, JSON)? What parameters are passed in the URL?
  
• Can an SSL endpoint with self-signed cert receive notifications?
• What if SSL cert of endpoint is expired?
• Are mail messages signed and if yes, how?
• How and when are messages actually persisted?
• The publish service isn't idempotent it seems?
  

PS: all based on reading the docs, must confess that I didn't actually test it

SSL Man-in-the-middle

Again a great "Security Now" podcast about SSL: how governments can sniff SSL traffic by enforcing Certificate Authorities to provide them with (intermediate CA) certificates. Based on this paper. Great story, recommended reading or listening!

Some things that I picked up:

• Different CA's can provide you with SSL certificate for same URL (or whatever)
• Internet Explorer (actually the Windows crypto) downloads extra CA's dynamically; so the list you see in IE can grow behind the scenes
• Firefox manages the list of trusted CA's itself
• There is no standard policy for when a CA is accepted by browser vendors
  
• The list of trusted CA's should be based on your geographical location
• Trusting a CA is somewhat equivalent to trusting a government
  
• Browser should provide (advanced) users with extra features to help them decide if CA certificate should be trusted or not

In my daytime job, SSL/TLS is used a lot for communication between IT systems within the corporate firewall or with business partners across the Internet. Low level configuration of SSL/TLS is often not supported:

• Configure single CA (or self-signed) cert to be trusted for specific outbound connection (e.g. when business partners have defined their "own CA")
• Different SSL client certificate per outbound connection
  
• Easy configuration revocation checks (OCSP etc); and checking if the revocation checks actually work
• Different timeout settings per connection
  
• Only accept SSL connections on specific interfaces