looking for windows (w2k) backup solution

[Brian Henning]

Joseph Mack NA3T wrote:
> On Tue, 17 Jan 2006, Brian Henning wrote:
> 
>> dd if=/dev/hdwhatever | gzip -9 | dd of=/path/to/storage/area
> 
> 
> hadn't thought about running gzip in here.
> 
> I ran nullfile, which puts a file DUMMY.??? in \winnt\system32. If you 
> ^C on nullfile (rather than letting it terminate normally) you're left 
> with a file in \winnt\system32. 

What?  I always copy the nullfile exe to C:\, run it there in a command 
shell, and it creates its temp file in the current working directory, 
C:\.  I've never seen it create its temp file in \%WINDOWS%\system32. 
Maybe it gets that working directory if you just double-click the exe 
instead of using a command shell; I've never tried it that way.  I would 
expect it to create the DUMMY.??? file in whatever working directory 
Explorer handed to it, in that case (which is my way of saying "blame 
Windows, not the program").

The abort behavior is useful for FAT filesystems.  Run nullfile until it 
error-terminates (when the FAT says "I can't make this file any 
bigger").  The DUMMY.??? stays, so you can run nullfile again 
iteratively until all the free space is used up, then del DUMMY.* . 
This, of course, isn't an advantage over the dd method.


> Deleting a file with a gui in a 
> directory with thousands of other files one day will lead to deleting 
> the wrong file. This doesn't seem like a good place for a temp file. A 
> better place would be one of the directories that gets emptied as part 
> of the disk cleanup in the explorer.

Sure.  No reason why my method wouldn't work by copying the exe to a 
system-cleaned temp directory.

> 
> Rob McCauley told me about temp directories and dd for windows, which 
> indeed I have with cygwin (how did anyone develope on windows before 
> cygwin?). I wrote a file of zeroes in c:\winnt\temp. Unlike linux, the 
> file is left there on termination, so you have to delete it by hand.
> 
> Using dd on windows seems more like familiar territory than running 
> nullfile.

Whatever makes you feel the most comfortable.  nullfile works without 
the cygwin layer (is there a native [i.e. not requiring cyg] dd for 
windows?  if so, wouldn't you still need cyg for the /dev/zero device?).

> 
>> It's left as an exercise for the reader to find out if block size on 
>> the reader dd makes any difference in the performance of the gzipping.
> 
> 
> block size doesn't make any difference with reading and writing. 
> Presumably everything's buffered, so possibly it will be buffered for 
> gzip too.
> 

...Which is why it was left as an exercise for the reader.  I had no 
idea one way or another, just a "hmm, wonder if..." and a lack of the 
inspiration to actually go find the answer. ;-)

~Brian

-- 
----------------
Brian A. Henning
strutmasters.com
336.597.2397x238
----------------