Re: Mike's correction of Joes correct Fred too

On 17/08/2024 03:27, Mike Terry wrote:

On 16/08/2024 19:50, olcott wrote:

On 8/16/2024 1:37 PM, Mike Terry wrote:

On 16/08/2024 12:59, olcott wrote:

On 8/16/2024 1:57 AM, Fred. Zwarts wrote:

Op 15.aug.2024 om 21:39 schreef olcott:
>
It is clear that olcott does not really read what I write. (Or is very short of memory.)
I never said such a thing.
I repeatedly told that the

>
*YOUR MISTAKE*

simulating HHH aborted when the simulated HHH had only one cycle to go.

That is WRONG. The outermost directly executed HHH aborts
as soon as it has seen enough of the emulated execution
trace to correctly predict that an unlimited execution
would never stop running.
>
*With abort as soon as you know*
*there is never one more cycle to go*
>
*MIKES CORRECTION OF YOUR MISTAKE*
On 8/14/2024 10:07 AM, Mike Terry wrote:
 > On 14/08/2024 08:43, joes wrote:
 >> HHH simulates DDD    enter the matrix
 >>    DDD calls HHH(DDD)    Fred: could be eliminated
 >>    HHH simulates DDD    second level
 >>      DDD calls HHH(DDD)    recursion detected
 >>    HHH aborts, returns    outside interference
 >>    DDD halts        voila
 >> HHH halts
 >
 > You're misunderstanding the scenario?  If your simulated
 > HHH aborts its simulation [line 5 above],
>
*THIS PART RIGHT HERE*
 > then the outer level H would have aborted its
 > identical simulation earlier. You know that, right?
>
 > [It's what people have been discussing
 > here endlessly for the last few months! :) ]
 >
 > So your trace is impossible...
 >

>
I supposed that I should be annoyed that you deliberately ignore my request to stop misrepresting my views and opinions.  You /know/ I don't agree with how you're misusing my words - but you do it anyway.
>

>
Both Joes and Fred seem to think that every HHH can wait for
the next one to abort and one of them will still eventually
abort.

 Fred above says that when HHH aborts simulated HHH, the simulation has only one more cycle to go before it terminates.  *HE DOES NOT SAY THAT HHH MUST WAIT ONE MORE CYCLE BEFORE ABORTING*.  And I'm pretty sure he doesn't think what you think he "seems to think".
 

>
Please try and explain to me exactly how your words did
not correct this error.

 Well first off - what you're challenging me to explain isn't something that either Fred or Joes were saying, so if you believed my words "corrected that error" then you had no justification in quoting me, because Fred and Joes /weren't making that error/.  This is just you not following the thread of conversation, or not understanding the meaning of what Fred/Joes are saying to you.  It would be like you saying "HHH correctly decides DDD" and I post a reply sending you to an atheist web site. When challenged I say "I thought you believed in God which is a mistake, so sending you to the web site would address that error."  [You see, it doesn't hang together...]
 Secondly, my quote above is pointing out why Joes' counterexample doesn't work.  It says that the /simulation/ of DDD by HHH never reaches DDD's final return e.g. because HHH *ABORTS* its simulation before that happens.  *NOTHING IN THERE ABOUT HHH WAITING ONE MORE CYCLE BEFORE ABORTING*.
 For the record, so you're not tempted to continue misrepresnting me:
 -  HHH /does/ abort its simulation of DDD before the simulation reaches DDD's final ret.
    (I'll go with Fred's "one cycle too early", for a suitable understanding of "cycle".
     The cycles aren't machine instructions, and each extra cycle we consider takes
     exponentially more machine instructions to simulate...  That's all ok.)
 -  From a /design/ perspective, coding a new HHH2 to be like HHH but waiting one more cycle
    achieves nothing because then its corresponding new DDD2 will also run for one more cycle
    before halting, compared with DDD.  So it remains the case that HHH2 aborts DDD2 one cycle
    before it will halt!
    So such a /design/ change does not help you.
    *I am not suggesting you redesign HHH to wait more cycles*
    *Neither Fred, Joes nor I believe that HHH waiting more cycles will fix*
    *your /design/ problem*.   [No design for HHH will work, as Linz proves.  Claiming
    one of your design decisions is "correct" because an alterntive fails makes no sense.]
 -  From a /run time/ perspective, yes, creating HHH2 to wait one more cycle enables it
    to correctly handle previous input DDD!  It will no longer abort too soon, so it will see
    DDD return and correctly decide "halts".  But Linz/Sipser don't contradict this -
    they both argue that HHH2 will decide /DDD2/ (not DDD) incorrectly.

Also I should have made clear here that if we are talking "Sipser quote" about HHH simulating input DDD, then Sipser's wording is "its simulated D would never stop running unless aborted".
In the case of your HHH/DDD, the simulation of DDD /would/ stop running if not aborted - it would stop in one more cycle.  This is demonstrated with HHH2 above, or with a "never abort" UTM.  THAT IS WHAT SIPSER MEANS BY HIS AGREEMENT TO THAT WORDING.  I.e. Sipser is talking "run-time" behaviour, not design-time change behaviour.  That's how I see it in any case.
So no way Sipser would become confused by your design-time coding change  which switches looking at input DDD to suddenly looking at DDD2 or DDD3 or DDDanythingelse.  His statement applies specifically to input DDD.
No way you'll understand any of that I guess...

 So what you're doing is confusing /design-time/ decisions that /you/ make, with /run-time/ decisions that HHH/HHH2 etc. make.  <Duh! PO slaps head in sudden understanding!!>  Also, you're calling different things the same name which would be confusing for anybody, but in your case it's worse, because you genuinely don't understand where different objects are involved.
  Mike.
 

>
If you keep insisting that I am wrong and fail to explain all
of the details of how I am wrong I will continue to assume that
it is your error of not paying close enough attention.

 You won't understand my explanation above in any case.  The point is that now you understand that you are misrepresenting my views - SO DON'T DO IT ANY MORE.