Re: Trouble translating a C++ data-structure.

["Dan'l Miller" <optikos@verizon.net>]

This Wikipedia article on aspect weavers is more lucid than the AOP article.  AOP's aspect weavers excessively emphasize a SQL database-esque join, as if that is the only game in town.  Alphard at CMU preceding the SEI did not approach the cross-cutting problem-space that way.  Alphard approached the topic far less like AOP and far more like multi-stage programming, which itself is an extrapolation of compiler technology beyond the typical 4:  source-code begets in-memory intermediate representation (e.g., AST) begets virtual processor's register-set & instructions begets mapping virtual processor to a particular physical processor's register-set & instructions.  Multi-stage programming (and Alphard) inject at least one more ‘begets’ generated source code and/or at least one more ‘begets’ transformations in intermediate representation.  Perhaps the former there and certainly the latter can perform what AOP does with aspect weaving but without emphasizing the SQL-database-esque joins.

> Mehdi Saada wrote:
> > I suspect also that between the cathedral and bazaar models of 
> > software development, you prefer the bazaar. Right ? 

As a software architect much of my professional career, I am cathedral in the extreme.

Paul Rubin wrote:
> From what I can tell, AOP is cathedral in the extreme. 

AOP is database-cathedral applied to imperative programming languages.  Imagine something like 3rd-normal form for source code.  Hence, AOP contorts the topic excessively, and perhaps too limiting.

Read all of Mary Shaw's work on Alphard from the 1970s.  It approaches the problem-space that AOP tries to solve from an entirely different perspective (which the INRIA OCaml community also pursue differently decades later):  little languages.  Alphard envisioned the big monolithic PL/I, Ada (then Green), C++ (not yet extant then), and so forth as a community of little languages.  When viewing it all that way, getting the community of little languages to conform to the specification of a big monolithic language strongly resembles the problem-space that AOP is (botching) aspect weavers to do, especially regarding cross-cutting concerns.

https://en.wikipedia.org/wiki/Aspect_weaver