The problem is well known, and a few strategies exists.

Cloudbees, for instance has a file system workaround and supports various types of parameters.

For me, using system environment properties (one could use environment properties as well) is simple and straightforward, and keeps a reasonably high level of security. Sure, if you have a lot of them (maybe API keys) configuring the application container can be tedious, but IMO is worth the effort.

The following is a simple Property object which chains a system variable into a Box and defaults to standard lift Props. The logic here is that, if you set a system property, that should take precedence. Another discussion of the problem here.

package code
package config

import net.liftweb._
import common._
import util._
import Helpers._

object Property {

  /**
   * Chain system property and Lift property
   * @param name key for the property to get
   * @return the value of the property if defined
   */
  def get(name: String): Box[String] = sys.props.get(name) orElse Props.get(name)

  def getInt(name: String): Box[Int] = get(name).map(toInt)
  def getInt(name: String, defVal: Int): Int = getInt(name) openOr defVal
  def getLong(name: String): Box[Long] = get(name).flatMap(asLong)
  def getLong(name: String, defVal: Long): Long = getLong(name) openOr defVal
  def getBool(name: String): Box[Boolean] = get(name).map(toBoolean)
  def getBool(name: String, defVal: Boolean): Boolean = getBool(name) openOr defVal
  def get(name: String, defVal: String): String = get(name) openOr defVal

}

Solution: download last year’s results and draw a histogram with R.

> res <- read.csv("Desktop/120325_semi_94.csv", header=TRUE, sep=";")
# convert time column to time
> temps <- strptime(res$Temps, format="%H:%M:%S")
# finally plot the histogram
> hist(temps, "mins", freq=TRUE)

Which gives this:

Indeed, this seems a quite fast race, but look at that nice spike around 2:00

Download the CSV results file or find more about the race