[mdlug] Realtek silent data corruption

[Mathew Enders]

On Mon, 2007-07-30 at 21:03 -0400, Raymond McLaughlin wrote:
> Michael Rudas wrote:
> > On 7/30/07, Michael Rudas wrote:
> > 
> >> Which also begs the question, "How did this driver get
> >> WHQL-certified?"  It shows real sloppiness on the part of MS and
> >> Realtek (I admit, no surprises there).
> > 
> > Oops... s/begs the question/raises the question... (tnx Carl).
> > 
> > First day with my new brain-- it takes a while for the neurons to knit.
> > 
> > -- Mikey
> 
> This also raises the question, at least among the ignorant like me, why
> did the error detecting/ correcting/ portion of the TCP stack, the crc -
> retry algorithms, etc, allow this? Sheer number of errors? Where the
> algorithms just over whelmed? Would a faster CPU make a difference? Is
> it just a matter that in a large enough population of bad packets some
> will, by chance, have the same crc check sum as the original?
> 

I think that after reading it that the data inside the packets was
corrupted not the packets themselves and that is how it passed the
TCP/IP check.  Some else asked the same question in the comments.