The present pipeline construction "create component" use case is written 
as if users will typically be creating the component algorithm at 
creation time, though the brief text of "Create Component Algorithm" 
does admit to the other possibility.

It may be a minor point, but I suggest we invert this. I would expect 
that a typical pipeline component will be a significant chunk of code 
(because the resulting pipeline will presumably be more efficient than 
if we componentize tiny bits of code). Such code will likely be 
developed and tested outside the pipeline construction framework.

So a concrete suggestion is to rename "Create Component Algorithm" to 
"Encapsulate Component Algorithm" or "Incorporate Component Algorithm".

There's a faint hint here of a library of algorithms waiting to be 
incorporated, but I don't think need a formal library with a library 
manager. Basically we can use the file system. I hope.

Regards,

-- Russell

_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data

Reply via email to