David,

Thanks for looking into this. I did read the documentation,
but obviously didn't quite get it.

> 
> Going back and reading the documentation a bit on this, it appears 
> that your assumption that adding SIDE_EFFECT results in the module 
> always runs is correct. However, the assumption that the downstream 
> modules assume that the output has changed is incorrect. 

I was assuming that the downstream modules would _detect_
the change, when they tried to lookup the new input in the cache.
I think I must not fully understand how the cache is used
or maybe how the executive does its dataflow analysis (or both!)

I was also assuming a broader definition of side effect,
wherein a module's inputs do not fully define its outputs.
The first couple of sentences of section 4.1 of the User's Guide
gave me this impression.

Cheers,
Joel

===============================================================
Joel Richardson, Ph.D.          [EMAIL PROTECTED]
The Jackson Laboratory          Phone:  (207) 288-6435
600 Main Street                 Fax:    (207) 288-6132
Bar Harbor, Maine 04609         URL:    www.informatics.jax.org
===============================================================

Reply via email to