Re: Concurrency in an RDB

NENASHI, Tegiri wrote:
>
> You have not showed this because we do not know a) what the compile is;
> b) why the compile in batches is different of the individual compile --
> why, when in the batch, one neeeds to compile more ? Your description
> is very general and non-understandable. What is this compile ? Is it
> that you cannot give an example of compile ? If this is so, then you
> are just playing with words.
>

The "compile" turns source code into object files ready for linking. It validates the symanics of the source code. A large project may take many minutes or hours to compile (ie, a long time to validate). You would want to compile batches of committed source code as the time to compile is much longer than the time to perform the mutative transaction. This is false economy *if* the committed source code often does not compile as it may be very expensive to determine which mutative transaction(s) should be nulled, let alone the issues of causality preservation.

I found David's example and reasoning very clear - are you not familiar with source code compilation? If that's the case, I hope this has helps clarify. Received on Wed Dec 20 2006 - 06:27:12 CET