[Fwd: Re: BAST in RAC]

  • From: Mladen Gogala <mgogala@xxxxxxxxxxxxxxxxxxxx>
  • To: "'Oracle-L (E-mail)'" <oracle-l@xxxxxxxxxxxxx>
  • Date: Thu, 23 Dec 2004 15:38:15 -0500

I am forwarding this rather heated exchange to the list, because I 
believe that the other
people may have something to say about this as well.

-------- Original Message --------
Subject:        Re: BAST in RAC
Date:   Thu, 23 Dec 2004 15:35:27 -0500
From:   Mladen Gogala <mgogala@xxxxxxxxxxxxxxxxxxxx>
To:     K Gopalakrishnan <kaygopal@xxxxxxxxx>

Comments inline

K Gopalakrishnan wrote:

>Interrupt is a generic word and it can be used anywhere in a relative
>context.  CPU hardware devices may have something called interrupt and
>I am least bothered about that while speaking in RAC related context. 
>In simple terms SAST/BAST are system traps sent to the owning

System traps are conditions on a CPU. There is a list of events.which 
trigger the processor trap
condition. Examples are "Illegal Instruction" (used in emulators), 
"Division by zero", "change mode kernel" and  "invalid address"  When 
trap condition arises, CPU activates trap handler (address is in a 
special register) which usually dispatches user handler, based on condition.
How does "process send a trap"? Trap is a SYNCHRONOUS event, and cannot 
be sent
from external sources. You are abusing VMS terminology again. On Unix 
systems, processes
do not send traps, (BTW, the original phrase was "enqueue AST") they 
send signals. So, your
trap mumbo-jumbo is some kind of signal or packet. Now were getting 
somewhere. If you
tell me exactly how is it implemented, we may actually untangle this 
unpleasant riddle.

>processes  and it is like an interrupt to that process. 

If you want to give a correct information, give it, if you don't, please 
refrain from mumbo-jumbo. Phrase "like an interrupt" is meaningless. On 
Unix, there are interrupts, traps, signals, files and sockets. There is 
nothing "like an interrupt". Sockets are "like interrupts", the same way 
I am like Mike.

> If you think
>that INTERRUPT should only be used to CPU/hardware decice context, you
>may want to go ahead and use that only for that purposes..

If you want to use it for fashion garments, that is fine with me, as 
long as there is
a clear definition. You did not provide one.

>>Your description of AST is what is scientifically known as mumbo-jumbo
>>(Oracle* Mumbo-Jumbo 10g) and makes no sense whatsoever. RAC processes work
>>over TCP sockets and do not "deliver interrupts". Processors and
>>hardware devices
>>deliver interrupts. Sockets deliver packets.  If you want to provide
>>internal information,
>>please do  it properly and not in a sloppy and misleading way using
>>incorrect and undefined terminology. This is  the same comment that I've
>>sent in private email to Steve Adams when
>>he was talking about "ASTs" in his internals book.
>What ever information I have provided is 100% correct in RAC related
>context. If some one else is relating that with VMS or punched cards
>then I can not help :)

Milk is not the same as Castrol Oil, regardless of your preferences. You 
can pour
it into your car, but it will not make it run another 3000 miles.  If 
you want to use
the terms like AST, "Trap" and "Interrup" for interprocess 
communication, that is fine
with me, as long as you define your terms precisely. If you don't do 
that, then your
"scientific explanation" is a meaningless mumbo-jumbe which does more 
harm then good.
You have not defined your terminology.

I will send a copy to the list, because I believe that other people have 
something to say about this

Mladen Gogala
Oracle DBA
Ext. 121

Mladen Gogala
Oracle DBA
Ext. 121


Other related posts: