Re: comments about Jigsaw design

Rob Mayoff writes:
 > I downloaded Jigsaw a couple of days ago and have been playing with it 
 > and looking at the code.  I'm very impressed with certain aspects of its 
 > design.  Specifically, it's method for tracking the meta-data and the 
 > mapping between resources and objects is excellent.  The configuration 
 > method (via forms) is also very well designed.
 > 
 > HOWEVER... ;^)
 > 
 > there are a couple of spelling errors in identifiers.  "Registry" is 
 > misspelled as "Registery" and "negotiated" is misspelled as "negociated". 
 >  Minor, but it's more frustrating to work with the code.

The first one will hopefully be fixed my next release, I'll fix the
second one too, but it's less important since the one I found was in
the private interface of NegotiatedResource.

 > My main thoughts concern the design of Resource filters.  To me, the 
 > Decorator pattern seems like the natural approach (like the way 
 > InputStream works).  But the Jigsaw implementation is more like the 
 > Strategy pattern.  See pp. 179-180 of "Design Patterns" (Gamma, et al) 
 > for a complete description.  It seems to me that the Decorator pattern 
 > would allow more powerful filters and be better encapsulation.

I always thought of it as the Decorator pattern, but I'll look more
closely.

 > Another one I wonder about is the way Resources are looked up.  In 
 > Jigsaw, the root container chases down the final target Resource in a 
 > loop, instead of just looking up the Resource for first component of the 
 > URL and passing the remaining components to that Resource.  Wouldn't the 
 > recursive-style lookup be better encapsulation?

Basically the way look up is performed - as far as I understand - is a
derecursivation of the recursive way of doing it. It has the same
effect, but coded in a slightly more efficient way.

Thanks for the comments,
Anselm.

Received on Tuesday, 16 July 1996 09:39:59 UTC