[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Dctc-general] Rollback resuming?
From: |
eric |
Subject: |
Re: [Dctc-general] Rollback resuming? |
Date: |
Fri, 20 Jun 2003 19:30:54 +0200 |
User-agent: |
KMail/1.5 |
Hi,
In fact, there is 3 main reasons to obtain a corrupted file after a download
(GDL or any other xfer type):
1) the original file is corrupted. This explanation is so easy that nobody
thinks of it. However, you will tell me that if the original file is
corrupted, one day, someone has had the uncorrupted file and I agree.
2) an uncorrected error during a transfer is another explanation. The checksum
used in ethernet is not perfect, it can detect a lot of errors but if you
check the specs, you will see one bit on 10^8 (ethernet standard) is not
corrected. It is very very small but it is one bit for 100,000,000 and a
700MB CD (mode 1) has 5,872,025,600 bits this means if you transfer a CD, you
may have 58 corrupted bits... Ouch!!! Hopefully, uncorrected bit does not
mean corrupted bit. On intel gigabit switch web page, they report they have
one corrupted bit for 10^12bits = one bit for 1,000,000,000,000. This means 1
corrupted bit for ~170 transferred CD. 170CD seems quite a lot for a user...
but not for the whole internet.
3) a buggy program or OS... I remember a program named hotline running on
win98 (first of the name) which sometimes, for an unknown reason, corrupts
the packet content but have a valid CRC (checksum) thus packets are assumed
to be ok by the remote client. It was not hotline fault (the same binary
works perfectly on NT) and the problem appears very randomly. It can also
came from stupid thing in a language like "I want the byte 1 of the file. VB
returns the first byte of the file but AFAIK, any other language will return
the second because all kernels (including windoz) has file first byte at the
position 0) [this looks stupid but it is a real case... direct connect works
like this :) ].
Eric