On Wed, Jul 29, 2009 at 5:51 AM, Jeremy Orlow<[email protected]> wrote:
> On Tue, Jul 28, 2009 at 9:30 PM, Peter Kasting <[email protected]>
> wrote:
>>
>> On Tue, Jul 28, 2009 at 9:23 PM, Mike Beltzner <[email protected]>
>> wrote:
>>>
>>> All we're doing at this point is preventing malicious applications from
>>> eating up disk, really.
>>
>> Yep, I agree (although that may no longer be true in a few years as web
>> apps grow in power and complexity).
>>>
>>> In the world of normal applications, you basically give them arbitrary
>>> permission to use your disk, but the good ones write some requirements ahead
>>> of time like "requires 200 MB free hard drive space" and warn you at install
>>> if you're below that.  Can we make the UI more like that, where you make a
>>> single trust decision up front?  Yes an app can lie, but normally-installed
>>> apps can lie too.  Can we provide enough ranking and feedback somewhere to
>>> make this decision easier on users?  For example, "57% of users chose to
>>> install <foo.com>, and gave it an average rating of 2.3 stars."
>>>
>>> Oooh, web of trust. There are some flaws. :)
>>> I do think the right answer here is to only get the user involved when
>>> the case seems pathological. Most uses of localStorage will be for "better
>>> than cookies," I suspect.
>>
>> One case I'm trying to prevent is getting separate requests, at different
>> times, from the same app.  You get some up-front query about desktop
>> shortcuts, and then a query five minutes later about using your camera, and
>> then a year later about going over 5 MB of storage, and so on.  Sucky.
>>  Really all I care about is an up-front "let this do whatever the heck it
>> wants" versus "no thanks".
>
> Another thing to consider is that, if our limits are per-origin (what
> most implementations use IIRC), a malicious attacker could easily use lots
> of host names (i.e. host1.bad-site.com through host10000000.bad-site.com) to
> still fill things up.
> I'm starting to wonder if some sort of web of trust or black list type
> solution is the only way to avoid users getting DOSed.

There was a nice paper at SOUPS about doing this for firewall rules:
http://cups.cs.cmu.edu/soups/2009/proceedings/a5-goecks.pdf

Slides: http://cups.cs.cmu.edu/soups/2009/slides/a5-goecks-post.ppt

> J

--~--~---------~--~----~------------~-------~--~----~
Chromium Developers mailing list: [email protected] 
View archives, change email options, or unsubscribe: 
    http://groups.google.com/group/chromium-dev
-~----------~----~----~----~------~----~------~--~---

Reply via email to