Re: Pattern Question
- From: raxitsheth2000@xxxxxxxxxxx
- Date: 27 Nov 2006 23:21:39 -0800
russel,
currently i m not having the book, but from your expression for the
plug-in , i didnt understand ur query ?
what you have described (plugin) is very simillar to "Programming to
Interface--not to implementation" and LSP
Registry somewhat imples that you are accessing existing instances of
types. If we are mainly building new instances does that violate that
pattern too.
I think if you are building new instances of *same type* then it
should not be the problem, But Beware, I am yet not sure (from ur
description), (this is also like Prog to I'face,LSP --I think registry
is more than these...)
--raxit
russell@xxxxxxxxxxxxx wrote:
I am having a discussion with someone in my office about two patterns.
plug-in and registry.
What we are building is a global definition of what type to use in a
particular implementation. This makes it sound like a good match to
plug-in but here is my problem. The way we are doing it allows for the
particular implementations to inherit from each other. In fact core
product types can be replaced via this definition for a particular
implmentation.
My concern with plug-in is that it in the Patterns book I am using
(Martin Fowler's Patterns of Enterprise Application Architecture) talks
about a plug-in pattern as Linking types at configuration time not
compile time. What we are doing does not support the abstraction of
no compile time linking. Is that big enough to though it out of the
pattern?
Registry somewhat imples that you are accessing existing instances of
types. If we are mainly building new instances does that violate that
pattern too.
What you would call the application component defined above?
.
- References:
- Pattern Question
- From: russell
- Pattern Question
- Prev by Date: Re: Half user config and half client prog implemented
- Next by Date: Re: Pattern Question
- Previous by thread: Pattern Question
- Next by thread: Re: Pattern Question
- Index(es):
Relevant Pages
|