[Qt-interest] Windows 7 and VirtualStore woes

[Constantin Makshin]

OK, I see. Anyway, I didn't (at least, didn't want to) say automatic  
updating is a no-no. I said it should be avoided when possible.

Obviously, when you have to update a program on many computers, updating  
each one manually is somewhat... problematic. And in your case you check  
that things aren't going to break after the update before clients get the  
new version of your application.

But doing an automatic update in uncontrollable environment with millions  
of different hardware-software combinations is risky because if even the  
new application works on your QA computer[s], it may not work on every  
user's computer. Besides that, not all users are technically advanced and  
they'll be very confused, if not scared, if your application suddenly  
begins to malfunction.

On Fri, 11 Sep 2009 10:53:26 +0400, Scott Aron Bloom  
<Scott.Bloom at sabgroup.com> wrote:
> No.. I mean, I put on a server the latest version and the url of where  
> to download it.
>
> My apps, on a regular basis, download the version, compare against it,  
> download & install it if its newer..
>
> My "push" is make the download available and rev the version.. Which  
> only happens after I pushed the new version and update app to the  
> internal QA server area.  Then test to make sure the updater works ok..  
> etc  Once that has QA sign off, I can push it to the server.
>
> Another example.. Try and run your Xbox or PS3 in "live" mode without  
> having the latest OS patches and updates..
>
> Automatic updates are a part of application development.  I have a class  
> that I have been using for about 3 years now, that downloads the latest  
> app, launches it, and shuts down the current app.. Its worked like a  
> charm for a long time now.
>
> It can be configured to handle "allow cancel", "always run" etc.. Then  
> when the apps get around to check.. they go.
>
> Scott
>
> -----Original Message-----
> From: qt-interest-bounces at trolltech.com  
> [mailto:qt-interest-bounces at trolltech.com] On Behalf Of Constantin  
> Makshin
> Sent: Thursday, September 10, 2009 5:20 PM
> To: Qt-interest
> Subject: Re: [Qt-interest] Windows 7 and VirtualStore woes
>
> Well, it looks like a misunderstanding of each other's words. Under
> automatic updates I mean completely unattended (e.g. launched by cron)
> ones.
>
> So, if under "I want to push out an update to my 2000 servers" you mean
> something like "I'm going to launch a script that'll update my 2000
> servers", that's a manual update in my terminology because you start the
> update process intentionally by yourself, it doesn't happen unexpectedly.
>
> On Fri, 11 Sep 2009 04:04:09 +0400, Scott Aron Bloom
> <Scott.Bloom at sabgroup.com> wrote:
>> What about the case of I want to push out an update to my 2000 servers?
>> Are you saying that I shouldn’t be able to do it automatically?
>>
>> Yeah.. I may break things.. And if it breaks it where I have to manually
>> update the 2000 servers.. Heck 10 servers.. Its gonna be a MAJOR MAJOR
>> main in the butt..  But that’s why I QA.. isn’t it?
>>
>> Scott
>>
>> -----Original Message-----
>> From: qt-interest-bounces at trolltech.com
>> [mailto:qt-interest-bounces at trolltech.com] On Behalf Of Constantin
>> Makshin
>> Sent: Thursday, September 10, 2009 4:50 PM
>> To: Qt-interest
>> Subject: Re: [Qt-interest] Windows 7 and VirtualStore woes
>>
>> True, but in that case automatic updates are even worse, IMHO. What if a
>> service begins to work incorrectly after an update?
>>
>> If you updated it manually, you know when the error occured and, if
>> necessary, send a bug report.
>>
>> But if it updates automatically, you're likely to miss the moment of the
>> update that caused the error. So you or somebody else work as usually  
>> and
>> suddenly "oops, things got broken".
>>
>> It's a particular case of "user doesn't want computer do what [s]he
>> didn't
>> ask for" case. When starting the update process, you understand that
>> something may get wrong and prepare for it (make backups, etc.). But  
>> with
>> automatic updates it'll be a lot harder to find the cause of the  
>> problem.
>>
>> That's my opinion, feel free to correct me. :)
>>
>> On Fri, 11 Sep 2009 03:04:34 +0400, Scott Aron Bloom
>> <Scott.Bloom at sabgroup.com> wrote:
>>> However,
>>>
>>> There are MANY MANY applications that HAVE NO user interaction, and are
>>> run on servers.  Services for example..
>>>
>>> Scott
>>> -----Original Message-----
>>> From: qt-interest-bounces at trolltech.com
>>> [mailto:qt-interest-bounces at trolltech.com] On Behalf Of Constantin
>>> Makshin
>>> Sent: Thursday, September 10, 2009 3:59 PM
>>> To: Qt-interest
>>> Subject: Re: [Qt-interest] Windows 7 and VirtualStore woes
>>>
>>> Nice article, thanks. But even taking it and its comments into account,
>>> I
>>> still don't recommend automatic updates, especially ones that cannot be
>>> cancelled. Nothing too bad will happen if the user ignores the update
>>> notification, but if the program suddenly changes its behavior (in any
>>> way) after an update, that'll confuse the user.
>>>
>>> The user doesn't want computer do what [s]he didn't ask for. E.g. when
>>> the
>>> user launches a web browser, [s]he expects that the browser'll only
>>> "browse the web" and not anything else.
>>>
>>> On Fri, 11 Sep 2009 01:19:59 +0400, Thiago Macieira
>>> <thiago.macieira at trolltech.com> wrote:
>>>> Em Quinta-feira 10. Setembro 2009, às 23.55.43, você escreveu:
>>>>> This may sound somewhat rude, but the standard way to solve
>>>>> self-updating
>>>>> problem is to not do it. It's better to implement notifications about
>>>>> new
>>>>> versions of your program. Something like "There's a newer version of
>>>>> Foo.
>>>>> Do you want to download it? <Yes/No>" and when the user clicks "Yes"
>>>>> button, open web page where [s]he can download new version of your
>>>>> program.
>>>>
>>>> That may sound like a sensible idea, but there's an obstacle to it:  
>>>> the
>>>> user.
>>>>
>>>> Please read 'The default answer to every dialog box is "Cancel"', a
>>>> blog
>>>> by
>>>> Microsoft's Raymond Chen:
>>>>
>>>> http://blogs.msdn.com/oldnewthing/archive/2003/09/01/54734.aspx

-- 
Constantin Makshin