On Dec 12, 2006, at 11:58 AM, Christos Zoulas wrote:
[ ...thread recovered... ]
>>> |> We are strictly talking about userland utilities being unable to
>>> |> handle large datasets because they are compiled in 32 bit mode.
>>> |
>>> | *You* might be strictly talking about userland utilities,  
>>> Christos;
>>> | claiming to speak for every poster to this thread is unlikely to
>>> | either be factually accurate or fair to other people.
>>>
>>> And everyone else in this thread, before you stepped in to mention
>>> how scalable the solaris kernel is, which we already know and  
>>> appreciate.
>>>
>>
>> Counterexamples include Message-id  
>> <274190120612100706jf397d37vcb1cd027024d9c5c@mail.gmail.com> which  
>> states: "I'd be interested in seeing numbers for the syscall  
>> latency, too."
>>
>> The latency of system calls involves a transition to kernel code  
>> by definition.
>> Would you retract your claim, please?
>
> This is not my message, I don't have a gmail account.

I didn't say it was your message; it was a message from someone else,  
earlier in the thread, which demonstrates that other people were in  
fact discussing the impact of 64-bit computing upon both userland  
*and* kernel performance.

> | Would you retract your claim, please?
>
> I did not make any claims about syscall latency.

"Your claim" was in reference to the statement that everyone else was  
talking strictly about userland utilities in this thread, which claim  
is trivially disproven by the gmail.com Message-Id quoted above,  
among others.

> Since you seem to selectively answer my questions, and the ones you  
> don't like you erase:

You know, I've done my best to be polite and to grant you credit that  
your position was the result of poor understanding rather than  
deliberate and willful dishonesty.  However, it is no longer possible  
for me to hold such an optimistic position in the face of your  
behavior, Christos.

When you refuse to acknowledge facts and you practice elision of key  
points (requoted above), and then accuse *me* of selectively  
answering questions, not only have you confirmed that I've given you  
too much credit in assuming you might be honest, but in addition you  
exhibit a pathetic brand of hypocrisy.

>> Now, let's focus on the discussion: While on the sparc/sparc64
>> model it probably made sense to have those utilities compiled in
>> 32 bit mode because of performance/compatibility, did it really
>> make sense on the x86/x86_64 environment? I think not.

Thanks for your opinion.  Oddly enough, many people think so highly  
of their own opinions that they choose to ignore facts which  
contradict their opinions.

If you want to accomplish something more useful than bandying words,  
Christos, why don't you try benchmarking 32-bit & 64-bit performance  
of NetBSD, or Solaris, or any other platform for a meaningful task,  
and post some real numbers rather than wasting time with opinions.

-- 
-Chuck