home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!destroyer!gatech!hubcap!fpst
- From: anik@crhc.uiuc.edu (Sadun Anik)
- Subject: Re: Synchonizing watches
- In-Reply-To: anik@crhc.uiuc.edu's message of 3 Sep 92 11:14:31
- Message-ID: <1992Sep4.174616.14440@hubcap.clemson.edu>
- Sender: fpst@hubcap.clemson.edu (Steve Stevenson)
- Nntp-Posting-Host: lyra.crhc.uiuc.edu
- Organization: Center for Reliable and High-Performance Computing
- References: <1992Sep3.123226.14205@hubcap.clemson.edu>
- <1992Sep3.172657.10929@hubcap.clemson.edu>
- Date: 4 Sep 92 12:40:39
- Approved: parallel@hubcap.clemson.edu
- Lines: 52
-
- In article <1992Sep3.172657.10929@hubcap.clemson.edu> anik@crhc.uiuc.edu (Sadun Anik) writes:
-
- In my reply I said:
- > OK I'll bite. It is trivial for A and B to figure out AB+BA. It is
- > also trivial for them to synchronize their watches assuming AB=BA (or
- > any preset assumtion on AB/BA). At this point their watches are
- > actually synchronized. You can not refute this with any communication
- > between A and B using this communication channel. My wild guess :) and
- > please tell me if this is wrong.
-
- Ed's comment.
- [I don't think that your assumptions are safe. Steve]
- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Maybe it was
- not clear in the previous paragraph but what I mean is A and B can
- assume a value for AB/BA and solve the problem. Because of the nature
- of the problem any assumtion is equally valid and would give you a
- solution. There is no safety problem here because this is just a
- choice for the value of a parameter. This is the case simply because
- as far as A and B are concerned their watches will be synchronized,
- and they are the only people which count. They can conduct time based
- experiments and their results and all their related communication
- (over the slow line) will verify this. Of course they are cutoff from
- external world so their experiments will be very limited.
-
- My understanding from the comment above is that when with an asumption
- like AB/BA=1/3 A and B synchronize their watches, sometime in the
- future they will realise that their their assumtion is not correct.
- As I said, if this is the case I would really like to know the
- algorithm. I don't have formal background in this topic so I
- can't prove what I said is correct.
-
- > If you are using a third person "C" to check if their watches are
- > actually synchronized using delays AC,CA,BC and CB, then the system you
- > are solving is perturbed so the previous solution doesn't need to be
- > valid. But this is irrelevant because this is not the system in the
- > problem. The solution is not also valid if you have another means of
- > communication between A and B.
-
- Even if you observe A and B and see that their watches are not
- "actualy" synchronized (A and B don't know anything about this
- observer) for A and B their watches are still synchronized. If
- the original question addressed the problem of synchronization as far
- as the observer is concerned, there shouldn't be any solution. This
- follows trivially from the solution above.
-
-
-
- --
- Sadun Anik, U of Illinois at Urbana-Champaign
- Center for Reliable and High-performance Computing
- e-mail: anik@crhc.uiuc.edu
-
-