[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: single-bit errors
From: |
Larry Jones |
Subject: |
Re: single-bit errors |
Date: |
Wed, 22 Nov 2000 14:13:24 -0500 (EST) |
Dennis Jones writes:
>
> We have noticed that some of our files in CVS are getting off-by-one bit =
> errors. For example, in a text file, an 'L' is converted into a 'K' and =
> a space into a '1' (two bits were modified here). This is not too bad =
> in text files because a differencing tool will usually detect the =
> problem, but binary files could be a more significant problem.
>
> Our developers are geographically separated, so we use pserver mode =
> where the server is accessible on the internet. Are there any built-in =
> CVS features or extra transmission verification methods we could use to =
> minimize the possibility of errors in CVS?
CVS pserver runs across TCP/IP which is supposed to be reliable (it has
checksums and other mechanisms to detect packet corruption). It's far
more likely that your corruption is being caused by some kind of a
hardware problem on your server: What kind of a system is it? If it's
a PC, does it have parity or ECC memory? What kind of disk is the
repository on? Normal IDE/ATA drives and controllers don't have any
integrity checking at all. If it's SCSI, make sure that you haven't
disabled parity checking on the SCSI bus. And I suppose you should make
sure that you haven't disabled the TCP/IP checksum checking, too.
-Larry Jones
I never get to do anything fun. -- Calvin