> So what happens when you check that the network is reachable before you 
> execute the operation, and then a network failure happens later during 
> the operation?  You've just moved the point of failure without actually 
> improving the user experience.

I agree that is a possible scenario, but as far as network intermittency is 
concerned, probably in the 0.01% range of occurrence, if not less.  The other 
99.99% of the time we would discover the problem before firing a doomed install 
request at a dead or unreachable address.  In the case you describe, we would 
figure out that the network *became* unreachable checking the error as you 
suggest.

Steve

-----Original Message-----
From: Shawn Walker [mailto:[email protected]] 
Sent: Wednesday, March 14, 2012 11:38 AM
To: Steve Sundstrom
Cc: Ervin Yan; [email protected]; Venkatesha M.G.
Subject: Re: [pkg-discuss] How to validate whether a publisher is reachable or 
not

On 03/14/12 11:22, Steve Sundstrom wrote:
> Checking before the operation would allow any script to verify
> whether the publisher was reachable and exit gracefully with a
> "Network is unreachable" message.  This is a very realistic
> scenario.

Yes, however, network operations can fail for a variety of reasons and 
at any time so attempting to determine if failure will happen before the 
operation is executed is not the correct solution.

The packaging system already provides appropriate error messaging for 
the failure cases.  If that failure messaging is inadequate or needs 
improvement, please feel free to file a request for enhancement through 
the appropriate support channel.

> Otherwise we get "Installing", "Failed", (figure it out) "Network is
> Unreachable" and a much more unpleasant user experience.  Our
> methodology is to do all possible to guarantee that an action will
> succeed before we attempt it.  It's the same as checking whether a
> packages dependencies are there instead of installing it to find out
> what is missing.

So what happens when you check that the network is reachable before you 
execute the operation, and then a network failure happens later during 
the operation?  You've just moved the point of failure without actually 
improving the user experience.

As I said above, the packaging system already provides the appropriate 
error messaging and handling for this scenario.  If you feel that it 
needs improvement, please provide specific feedback as to how.

Thanks,
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to