[Info-vax] yet another sys$qiow question

[Stephen Hoffman]

On 2015-08-22 19:48:03 +0000, Bob Gezelter said:

> Au contraire. $CANCEL indeed cancels IO.

$cancel can be a bag of hurt.  With some devices it works, with some 
others sometimes not so much.   Whether those problematic cases are 
because the device can't cancel the I/O, or due to the timing of the 
$cancel, or due to driver bugs?  Sure, toss a $cancel at the device, 
but expect it might just ignore you.  Or it might not.  Best to assume 
that the I/O operation is either going to arrive or error out, and 
maybe $cancel shortens the wait.  Distributed designs built on 
protocols with guaranteed delivery and back-pressure — have had more 
than a little "fun" with DECnet here — can be surprisingly nasty to 
deal with.  $cancel "just" adds more to the state machine table.



-- 
Pure Personal Opinion | HoffmanLabs LLC