[ros-dev] Integration of ultracopier to reactOS

Love Nystrom love.nystrom at gmail.com
Mon Nov 2 16:32:14 CET 2009


Wesley Parish wrote:
> Isn't the NT code base supposed to have multi-queued IO?  to stop such 
> problems?  At least that was the NT-zealots' boast during the NT-vs-OS/2 days 
>   
I'd say it's one thing what the system has ability to do do, and another 
thing
entirely what Explorer does with it. Not knowing it's source, I can only say
"by the fruit shall the tree be known".
> Shouldn't the routine to copy hold another IO queue in hand "just in case", 
> and hand the copying over to it as soon as it hits such a snag?
Precisely. A fault list or whatever you want to call it.
This could then be post-processed at the option of the user.

> Mind you, I could think of additional uses for a multi-queued IO - having a 
> anti-malware program kibitzing on the copying, and stalling anything that's 
> questionable, while handing the copying over to Yet Another Copying IO Queue.
>   
Bear in mind that there's an inherent danger in exposing the shell's 
file copying mechanism.
While it may seem a convenient way to tie in a malware scanner, it could 
be used to conveniently
inject a payload in every file the shell copies if a "plug-in" gets 
write access to the file data.
Because of this risk any "plug-in" would have to be handed just a copy 
of the data, which would
lead to terrible inefficiency due to all that redundant data copying.

Rock on // Love




More information about the Ros-dev mailing list