I think the safest thing to do is probably to check the update url for
updates and have it just return some basic information like update
available yes/no. If the gadget gets a yes response it can popup a
predefined message to the user "There is a new version available,
visit our webpage to download it now" and then redirect him to the
gadgets webpage which is hard-coded in the gadget itself. You could
add the current version as a get parameter here and greet the user
with the update message if he in fact needs a new version.

Benjamin

On Sep 24, 12:17 pm, eharman <[EMAIL PROTECTED]> wrote:
> Thanks for your response, this is most interesting. I didn't think
> there was anything special about framework.operUrl, and since we're
> trying to be backwards compatible to 5.1 I think we'll continue using
> the activexobject.
>
> OK, so that's a lot to think about. Firstly you advise using a secure
> connection whenever posting data; happily we're already doing this.
> The gets are using non-secure connections, but (with the exception of
> the download URL) all responses are verified before parsing.
>
> Secondly all asynch responses are XML, not JSON, so no danger of eval-
> ing something nasty. The download URL is the only instance where un-
> audited response data is given directly to the user.
>
> Thirdly, there's the question of the update mechanism. At the mo it
> looks like this:
> 1. Asynch request for version info + update URL to new .gg file (not
> secure)
> 2. Gadget displays upgrade message and "download" link
> 3. User clicks link, which uses WScript.Shell to run the URL in the
> default browser (not secure)
> 4. User is prompted to save/run by the browser, following the normal
> browser download flow.
>
> If I'm understanding correctly, an attacker might intercept the asynch
> version/URL request (#1), and modify the response to include a bogus
> update URL. I totally agree that we should not be providing a direct
> URL to the .gg file; thanks to this thread I now have external support
> to convince people to provide a download *page* instead. Much better
> usability if nothing else, as we can give instructions on what to do
> if the install doesn't work.
>
> Even given that change, I suppose technically our asynch XML response
> would still be vulnerable to attack: someone could still modify it to
> a URL of their choosing. I'm guessing we should also make this request
> secure, just to be on the safe side?
>
> Thanks for the clarification on how MITM attacks work, and that simple
> links to web pages are low in risk. I think other than the two fixes
> for the upgrade mechanism, it sounds like we're pretty clear. Am I
> missing anything that you can tell? BTW sorry I can't send the gadget
> itself; it hasn't been released yet and blah legal blah.
>
> Elaine
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"Google Desktop Developer Group" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/Google-Desktop-Developer?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to