Hi,

I was reading in Expert F#, Chapter 13, the example on asynchronous image processing using the Async workflow. Much the same reasoning below can be applied to canonical workflow examples, as in [link:blogs.msdn.com]

The workflow in the Expert F# example reads an image file asynchronously, does some processing, and writes it back to disk asynchronously. Many of these workflows are started in parallel using Async.Parallel. The text then explains that this speeds up the processing time.

However, I would assume that this speed up is _only_ due to the fact that multiple read-process-write process are started in parallell (i.e. the program reads multiple images). The fact that the reading and writing of files occurs asynchronously _within one of the threads_ does nothing to speed things up, as the rest of the computation just waits until the asynchronous read or write returns. These steps might as well execute synchronously, yielding the same overall result. (At most, this asynchronous behavior gives you some

So, it seems that the ReadAsync & friends methods are not very useful in this example - they would be useful if some Future container would be returned, and if the calling thread would not be blocked unti it's result is actually necessary.

What am I missing?

Kurt

PS I'm not debating the overall usefullness of workflows, or Async workflows. It just seems to me that this particular (and often used) example has some serious flaws.