W3C home > Mailing lists > Public > public-wsc-wg@w3.org > January 2008

Re: ISSUE-131 (Code outside browser): Executing code outside of browser in 8.3.2.3 is vague / scary [All]

From: Anil Saldhana <Anil.Saldhana@redhat.com>
Date: Wed, 02 Jan 2008 12:01:15 -0600
Message-ID: <477BD16B.1000902@redhat.com>
To: Ian Fette <ifette@google.com>
CC: michael.mccormick@wellsfargo.com, public-wsc-wg@w3.org

Since I have never written a browser or a plugin, I cannot claim to be
an expert in these two areas. :)  But I will still write what I think.

It should not be difficult for a plugin writer to convey to the browser
that on a given platform (OS etc), the plugin will execute outside the
browser context. Before the plugin kicks in, it can reflect this on to
user (with a consent dialog) such that the user can always blame himself
in the end, for agreeing to execute the plugin. :)

Now to the question of the browser popping up a consent dialog each time
a pdf is opened, the consent dialog can certainly have a "Remember my
decision for this plugin xxx" such that you have a seamless pdf viewing
each time.

All I wish for is when a plugin tries to do anything different from what
my expectations are (outside the browser context or upgrade
automatically), I am told about it. In the case of the flash plugin
upgrading itself, it is not very difficult for them to add a notice
inside the flash movie chrome such as:
"The execution of this flash movie requires that you upgrade your flash
plugin version to v10. We can do it automatically for you. Press Yes if
you agree. No to cancel".

Ian Fette wrote:
> I think there are a lot of things that suck, that could probably be
> eventually fixed if browser vendors took a stand. Plugins are one of them,
> user agent strings are another (have you seen the iPhone user-agent string?
> "Mozilla/5.0 (iPhone; U; CPU like Mac OS X; en) AppleWebKit/420+ (KHTML,
> like Gecko) Version/3.0 Mobile/1A542a Safari/419.3" This is absolutely nuts,
> and is huge overhead that is being transmitted everywhere. Apple could take
> a stand and say "This is stupid, we're just going to say "iPhone -
> AppleWebKit", but then tons of sites would break due to crap code doing
> (often unnecessary) UA checking in a poor manner. Could apple change these
> things, like warning on plugins doing wacky IPC, crazy UA strings, etc?
> Sure. Would the responsible parties eventually fix it? Probably. But who
> suffers in the meantime? The user.
> 
> I'm not sure I like requiring things that cause the user to suffer, because
> it means that either the user is going to suffer (bad), or the browsers
> aren't going to implement the change because it would cause the user to
> suffer (also bad).
> 
> If we have problems with things like NPAPI, or plugin rights, those should
> be addressed somewhere, but I don't think that trying to slip them in
> WSC-XIT is the appropriate place to do so.
> 
> On Jan 2, 2008 9:23 AM, Anil Saldhana <Anil.Saldhana@redhat.com> wrote:
> 
>> On many platforms, the acrobat reader opens the PDF within the browser
>> chrome (same tab on firefox).
>>
>> I think this is an important requirement from an user's perspective that
>> they be notified when a plugin tries to execute things outside the
>> contract established between the user and the browser. I understand that
>> it is going to be extremely hard in getting it right. But until the
>> browser vendors/implementers raise a red flag on this, I support the
>> retention of this bullet. :)
>>
>> I did recently mention the case of the Adobe flash plugin automatically
>> upgrading itself due to the new Flex software intentions. The user has
>> no control over the upgrade. Just because a flash movie requires an
>> upgraded plugin, does not mean the user has no say in consenting to the
>> plugin upgrade. :)
>>
>> Ian Fette wrote:
>>> As per our 12/12 meeting, I am proposing removing the third bullet under
>>> 8.3.2 - "Web user agents MUST inform the user and request consent when
>> web
>>> content attempts to install or execute software outside of the browser
>>> environment". There are many things that make this hard / impossible to
>> get
>>> right, and even harder to actually get the intended effect without being
>>> totally annoying.
>>>
>>> For instance, when you load a PDF, Acrobat Reader is launched outside of
>> the
>>> browser context. Yet I don't really want a dialog box every time I
>> browse to
>>> a PDF, I just want to see the PDF. Same thing when I click on a mailto:
>> link
>>> - it's going to get shell executed, and software (my MUA) is going to
>> run
>>> outside the browser. Or if there's an embedded video that causes the
>> windows
>>> mediaplayer plugin to do some funky COM stuff outside of the browser -
>>> again, I really don't want dialog boxes here. I understand the intent
>> and
>>> think it's probably a good one, but it's really hard to actually get it
>>> right in words, and I think it's something that browsers are doing
>> pretty
>>> well anyways.
>>>
>>> I'm not going to rehash everything in this email, please see the 12/12
>> notes
>>> for a full review of the conversation (
>>> http://www.w3.org/2007/12/12-wsc-minutes.html ). In that meeting, I said
>> I
>>> would email back on this issue and propose that the best way to resolve
>> it
>>> is to simply remove the bullet point, unless anyone feels strongly about
>> it.
>>> If you do feel strongly about it, then please come up with some
>> alternate
>>> text.
>>>
>>> Thanks,
>>> Ian
>>>
>>> On Nov 6, 2007 8:36 AM, <michael.mccormick@wellsfargo.com> wrote:
>>>
>>>> The "install" part is very important, but the "execute" part is a
>> rabbit
>>>> hole we probably don't want to go down.
>>>>
>>>> For example, when I point IE at a resource of MIME type ms/xls, Excel
>>>> launches outside the browser as a helper app.  It would be annoying if
>> I
>>>> got constant warning messages every time I pull up a XLS, PDF, etc.
>>>> Constant warnings = ignored warnings.
>>>>
>>>> I do want to be warned when a page tries to install a plugin like
>>>> Acroread, but not every time that plugin runs.  Same for helpers,
>>>> toolbars, extensions, ActiveX controls, etc.
>>>>
>>>> -----Original Message-----
>>>> From: public-wsc-wg-request@w3.org [mailto:public-wsc-wg-request@w3.org
>> ]
>>>> On Behalf Of Web Security Context Working Group Issue Tracker
>>>> Sent: Tuesday, November 06, 2007 9:50 AM
>>>> To: public-wsc-wg@w3.org
>>>> Subject: ISSUE-131 (Code outside browser): Executing code outside of
>>>> browser in 8.3.2.3 is vague / scary [All]
>>>>
>>>>
>>>>
>>>> ISSUE-131 (Code outside browser): Executing code outside of browser in
>>>> 8.3.2.3 is vague / scary [All]
>>>>
>>>> http://www.w3.org/2006/WSC/track/issues/
>>>>
>>>> Raised by: Ian Fette
>>>> On product: All
>>>>
>>>> 8.3.2.3 says "Web user agents MUST inform the user and request consent
>>>> when web content attempts to install or execute software outside of the
>>>> browser environment."
>>>>
>>>> This is a bit vague and probably not what we intend. For instance, when
>>>> you navigate to a PDF on a browser using Acrobat Reader w/NPAPI plugin,
>>>> what happens is that there is a plugin running in the browser, and then
>>>> Acrobat Reader launches in the browser, and there's a ton of IPC
>> between
>>>> the plugin and Reader running in the background (which is doing the
>>>> heavy lifting). This is executing software outside of the browser
>>>> environment, yet I don't think this is really what we were intending to
>>>> warn users about. At least, I will scream if I get a popup every time I
>>>> navigate to a PDF. Seriously.
>>
> 

-- 
Anil Saldhana
Project/Technical Lead,
JBoss Security & Identity Management
JBoss, A division of Red Hat Inc.
http://labs.jboss.com/portal/jbosssecurity/
Received on Wednesday, 2 January 2008 18:01:27 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:14:20 UTC