Method Annotations

Stephane Ducasse ducasse at iam.unibe.ch
Thu Aug 21 07:49:29 UTC 2003


Hi,

I did not look at the way VW is doing it the menu registration based on 
pragma and may I should. If you look at the meta-data proposition for 
Java. You will see that quickly you will have a lot of builder specific 
tags and slowly but surely this will be the mess. I was discussing with 
an OTI well placed guy and he asked me if they would finally succeed to 
kill the language. Surely not because people like gory details and 
exception. But he said annotation this is reflection for the poor in 
substance.

Now I'm not against annotation. We were thinking a lot about 
annotations here too and exchanged some emails with andreas about one 
year ago. I would like to have annotation so that the system could use 
to warn the user. Example the new method compilation. I do not want to 
have this stuff hardcoded somewhere in the compiler why only #new and 
not #class, #name. However I do not want to have annotation that are 
suddenly changing the execution of the program because annotations will 
then mix meta and base level.

Roel would like to have annotations but for any parse tree element so 
that we could build clever tree transformation, AOP support, type 
inference....



About the process. Avi I do not like too much the idea of simple stuff 
then extend it. This is good when you do not know the final answer. I 
can tell you that if the SWT would have not damaged and sunk itself 
with a certain amount of funny arrogance I would prefer a lot a good 
declarative model build based on years of experience then just a try 
and an error system (even by smart people)  Understand here that I like 
your effort around package info but I'm sad that Joseph and you cannot 
collaborate
but this is not my business I just see a fact.) But this is not the 
point of my point.

Stef






More information about the Squeak-dev mailing list