CVS difference for ais/ai-00260.txt

Differences between 1.4 and version 1.5
Log of other versions for file ais/ai-00260.txt

--- ais/ai-00260.txt	2001/06/03 04:21:30	1.4
+++ ais/ai-00260.txt	2001/07/14 00:01:48	1.5
@@ -667,3 +667,75 @@
 Let me know what you think about this proposal ? Do you need something else ?
 
 ****************************************************************
+
+From: Pascal Obry
+Sent: Wednesday, June 06, 2001 3:54 AM
+
+Randy Brukardt writes:
+ > Attached is the AI that resulted from your proposal and the discussion about
+ > it at the Leuven ARG meeting. (Note that this discussion was for intent
+ > only, so the proposal is a draft only.) Any comments are welcome.
+
+I had a look and it seems fine with me. The feature I was missing is indeed
+available with this proposal. The wording is quite better than mine :)
+
+I have only one question, why choose 'Tag_Write and 'Tag_Read instead of
+'External_Tag'Read and 'External_Tag'Write. Here we are talking about external
+tag. The way to change the external tag represenation is to use External_Tag
+attribute. It seems quite straitforward to just support two new attributes on
+External_Tag to have finer control on the external tag representation.
+
+Of course this is a minor issue and I will be able to live with Tag_Write and
+Tag_Read.
+
+****************************************************************
+
+From: Randy Brukardt
+Sent: Wednesday, June 06, 2001 6:49 PM
+
+> I had a look and it seems fine with me. The feature I was missing is indeed
+> available with this proposal. The wording is quite better than mine :)
+
+Thanks for the complement.
+
+> I have only one question, why choose 'Tag_Write and 'Tag_Read instead of
+> 'External_Tag'Read and 'External_Tag'Write. Here we are talking about
+> external tag. The way to change the external tag representation is to use
+> External_Tag attribute. It seems quite straitforward to just support two
+> new attributes on External_Tag to have finer control on the external tag
+> representation.
+
+Two reasons: 'External_Tag is a value, not a type. It seems very strange to
+associate a subprogram with a value. Secondly, this would require a new
+mechanism in compilers: it would be necessary to recognize two attributes at
+once when implementing an attribute definition clause. Currently, compilers
+simply look at the trailing attribute (which, in most implementations, is at
+the top of the tree), and then handle the prefix appropriately. While
+'Tag_Read and 'Tag_Write use the same mechanism as for the other stream
+attributes, which is well understood.
+
+Note that there is nothing preventing implementors from implementing this
+tomorrow (because they are allowed to define implementation-defined
+attributes).
+
+****************************************************************
+
+From: Pascal Leroy
+Sent: Thursday, June 06, 2001 2:57 AM
+
+On the one hand I agree that somehow 'External_Tag'Read looks nicer than
+'Tag_Read.
+
+On the other hand if you look at the syntax for attribute definition clauses
+(RM95 13.1(3, 5)) you see that it would be entirely new syntax.  Therefore,
+it would require sizeable changes in the compilers for purely aesthetic
+reasons.  Not worth it.  It's much simpler to come up with new attributes.
+
+****************************************************************
+
+From: Pascal Obry
+Sent: Thursday, June 07, 2001 1:03 PM
+
+Right.
+
+****************************************************************

Questions? Ask the ACAA Technical Agent