On Dec 3, 2010, at 2:17 AM, Alvaro Herrera wrote:
> Excerpts from Robert Haas's message of jue dic 02 21:10:48 -0300 2010:
>> On Thu, Dec 2, 2010 at 3:43 PM, Alvaro Herrera
>> <alvherre@commandprompt.com> wrote:
>>> Excerpts from Robert Haas's message of jue dic 02 17:27:01 -0300 2010:
>
>>> Yeah, the Oracle system is a lot more complex than SQL Server's, but I
>>> was only talking about the latter, for which see here:
>>>
>>> http://www.databasejournal.com/features/mssql/article.php/3635426/SYNONYM-in-SQL-Server-2005.htm
>>
>> Well, that seems primarily designed to cut down on three and four part
>> names. We don't have that problem anyway.
>
> Right. (My point here is that SQL Server is not a good guidance on what
> the synonym system should do.)
>
>>>> The list of objects for which they support synonyms is also
>>>> interesting.
>>>
>>> The bit that allows a synonym to reference another synonym seems like
>>> worth considering further (either reject them altogether, or have some
>>> way to deal with possible cycles).
>>
>> It's pretty trivial to do cycle-detection at runtime.
>
> No disagreement on that, but something needs to be decided.
I don't think it makes sense to allow synonyms for synonyms. It would make
resolution code slower, and I don't see any situation where they make sense.
The original proposal didn't mention them, but limited the list of initially
supported objects to those to tables, views and sequences, implicitly
excluding synonyms referring to another synonyms.
--
Alexey Klyukin http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc