jmr wrote:
> Comments inline:
>
> JR
>
> Brock Pytlik wrote:
>   
>> Padraig O'Briain wrote:
>>   
>>     
>>> On 12/01/08 19:28, Brock Pytlik wrote:
>>> [snip]
>>>     
>>>       
>>>> If the user changes the refresh period, to be shorter than the time 
>>>> left to the next check under the previous policy, shouldn't the 
>>>> timeout be changed or another call added to the idle list with a 
>>>> shorter timeout? For example, if I set the refresh period to a day, 
>>>> wait till one minute after the refresh is happened (so the next 
>>>> refresh is set for 24 hours) then change the refresh period to be 
>>>> every hour, I shouldn't have to wait 24 hours before the hourly 
>>>> refresh period takes effect.
>>>>
>>>>
>>>>       
>>>>         
>>> This cannot happen. The shortest refresh period is 1 day and the 
>>> maximum that we wait for is 1 day; see line 352.
>>>     
>>>       
>> This seems like code that's written to be broken in the future when 
>> specs change. I see no reason that the shortest refresh period should be 
>> a day going future. Personally, I might like to have it be an hour, or 
>> 15 mins if I'm really excited about trying new bits. I also happen to 
>> think the maximum waiting of a day is broken as well, but I'm not 
>> interested in fighting that battle now. I do think that to put code in 
>> that's explicitly dependent on never having a refresh period shorter 
>> than 1 day begs to introduce bugs in the future.
>>   
>>     
> There is a cost to doing the check and 1 day seems a reasonable refresh 
> period, if we get more folks asking for this we can add in a shorter 
> refresh period. Of course if you are particularly keen to check for an 
> update you can just run the Update Manager directly from 
> System->Administration->Update Manager
>   
I think this misses the issue. The code being suggested makes a 
fundamental assumption that refresh period will never be shorter than 1 
day. I think that's a bad thing to introduce to the code unless its 
absolutely necessary. So far, I haven't heard that a) it's impossible to 
do it any other way or b) it's ETOOHARD to do it any other way. I 
continue to ask, why introduce code that it's easy to see may introduce 
bugs in the future.
>>> [snip]
>> To be clear, let's try a hypothetical example and see if I understand 
>> what you're saying isn't broken behavior.
>> On Dec 2nd, I boot my system and get notified that there's an update to 
>> random_package. I choose not to install that package at that time 
>> because it's not critical and I have work to do, so I cancel UM. I leave 
>> my system running all the time in general, so I don't reboot or log out. 
>> Now, Dec. 3 comes along and a new OS build is released that I do care 
>> about because it fixes bug XYZ, but I don't know it's been released 
>> because I'm not subscribed to the right email aliases (like most of our 
>> users). Because I have had UMN notify me once already,  I think that UMN 
>> will notify me when any new updates are available, let alone huge 
>> changes like a new OS build. In fact, I will never receive an 
>> notification of updates available again until I reboot or re-login, 
>> which could be months given the general stability levels Solaris has.
>>
>> How is this not broken?
>>   
>>     
> Ok - the use case is clear, I for one would like to have the UMN shut 
> down after the recheck so its one less thing running on my system, and I 
> reboot daily as a mobile user so this is not an issue.
>
> Lets add in another gconf key for UMN to allow this behavior to be 
> controlled:
>
> gconf key: shutdown_after_update_icon_clicked - default it should be set 
> to False, so it will not shutdown after user clicks on the UM 
> Notification icon.
>
> Setting this key to True will result in the UMN shutting down after the 
> user has been notified of an update and clicks on the UpdateManager 
> Notification icon. By default the UMN continues to run checking for 
> further updates after the icon has been clicked.
>   
That's fine with me.

Brock
>   
[snip]
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to