Well, the large file offset stuff in smbclient seems to work

Michael B. Allen miallen at eskimo.com
Sun Feb 16 05:48:13 GMT 2003

On Sat, 15 Feb 2003 23:04:29 -0500
"John E. Malmberg" <wb8tyw at qsl.net> wrote:

> Michael B. Allen wrote:
> >>
> >>No.  Strange effects can happen at many different file sizes.  If you do 
> >>not test it, you do not know that it works.
> >  
> > Can you give me a specific example? I've written a client and I never
> > tested it past 5-6GB. You have me worried now :-/
> There may not be a problem in your client.
> But problems may show up in file systems and the support C library 
> calls.  In older systems, bits were precious, so there may be many 
> fields that do not have enough, and now backwards compatability may be 
> showing it's age.  Sometimes it is found in a device driver that because 
> at the time a 1GB disk was unimaginable, that the bits above there were 
> used for flags.
> Some algorithms are sound but do not scale well, hence the unexplained 
> slowdowns.
> Every 4 bit nybble barrier can be an issue, and the signed/unsigned 
> usage may also be an issue.  The granualarity of blocks in the file system.
> Once you get past 4GB, I would expect the next hiccup may be at the 1TB 
> level and then every power of 2 beyond that.

Ok, so you're citing unforseen problems with how libraries, drivers,
hardware, etc handle the 64 bit type as opposed to some issue known to
the protocol like some bits getting trampled if the field isn't aligned
with the planets or some such.

> How many people are dealing with files larger than 4G on a regular basis?

Not many.

> You can not test every thing though. :-)

Yes, even over loopback the process is a too tedious :-)


A  program should be written to model the concepts of the task it
performs rather than the physical world or a process because this
maximizes  the  potential  for it to be applied to tasks that are
conceptually  similar and, more important, to tasks that have not
yet been conceived. 

More information about the samba-technical mailing list