World: r3wp

so in your opinion, plug-in should not be allowed to write even in 
sandbox dir?

Scripts that aren't cryptographically signed (anonymous scripts) 
would be limited in what they can do on your system.

my question was a bit different, though. Let's imagine your proposed 
secure way, so I can use it for some app, to sync some data to user. 
The problem for me is the sandbox placement - in system Temp dir, 
which can be purged via control panel. I am not sure I like it, if 
there is possibility I could loose my synced data. Or am I missing 
something?

Your last reaction to when I brought this up:

OK - one thing is clear now - "What would you let your worst enemy 
do with your computer?" should be a saying for Rebol plug-in .... 
now just how to represent it ...

My suggestion was just for anonymous scripts. Signed scripts could 
get a sandbox, probably somewhere under %appdata%.

Did it take long time for you to dig it up? :-)

I just had to look for the last point in the history where I posted. 
Security seems to be a recurring theme in my posts. :)

btw - what do you mean by "should not get a persistent sandbox at 
all" - do you mean it should not be allowed to write to temp at ll, 
just use memory? Or just by anonymous you mean randomly generated 
"anonymous" directory somewhere in Temp directory?

Isnt' this a bit over the top?  What about cookies ?

No, just use memory. No cookies except browser cookies - there should 
be a way to access them, similar to how do-browser lets you access 
JavaScript. Perhaps a wrapper function around do-browser could work 
for now.

although some storage for graphics-heavy things would be nice.

Brain - are you suggesting so tight security for us rebollers only, 
or do also other plug-ins use such limited environment? (although 
I am starting to understand, that there should be NO way of how to 
harm your computer, or it will be regarded - unsecure)

I don't want there to be any need for people to make a REBOL-blocker 
like FlashBlock or NoScript.

(both of which I use,  btw)

i would not go oncrypto only but also on "allow website", like noscript

Yes, we need better security requestors too.

yes, and remembering such things. maybe asking on start to allow 
access to last session. if denied sandbox is cleared.

that is one extra click

Volker, I agree that some graphics-heavy scripts could use local 
storage. Those scripts could easily be signed though.

There should also be a way to provide access to the browser's objects. 
The browser already caches those, and that cache is managed by code 
that the user is already trusting.

signing needs keys. then we need a free registry if we want all the 
newcomers to have fun.

and allowing access based on url is IMHO the most natural way.

Actually, there is a lot that you can do even within those restrictions. 
Just look at Flash.

but there could be done more :)

speciallly when things start. later onecan optimize, but thefirst 
 protoype-bitmaps can be large.

and if you cancel "reuse last session" and check "forever" you are 
pretty much anonymous.

Well then, the question you should consider when thinking about newbies 
is this: What would you let your worst enemy do with your computer? 
This is the web. We aren't talking about saints here - we are talking 
about people who use baner ads to install spyware.

I did ask. cant see doors for banner-adds nor spyware.

Banner ads are on web pages. You can make banner ads with Flash, 
and that is less dangerous than the current plugin.

files are a risk to privacy if they cant be blocked. that reuse-question 
does this. and they can be prepared to be run, eg called *.exe and 
hoping the user some day clicks on them. so i suggest a wrapper, 
maybe store everything as rebol[]#{stuff} or in a single zip or something.

i thought we talk about local storage. what has that to do with banner-addds?

wouldn't e.g. local storage limited to say 20 MB be sufficient "security"? 
Hmm, should read that Flash security doc first probably :-)

As I've mentioned here before, there many nasty things you can do 
with the present plugin and I don't want to make suggestions on a 
web-public group. Go private if you want some ideas - I trust you 
not to misuse them.

I read the Flash security doc, and it has many good ideas. I'm still 
a little iffy about it providing cookies to anonymous scripts without 
providing a management interface - that's why I still use FlashBlock.

how can cookie harm you?

I mean Flash cookies - browser cookies do have a management interface.

Cookies can be used to track your movements, and can be used as persistent 
distributed storage.

I trust you gurus for proper security concerns, however let's not 
forget us - the end users, where over complication does not work. 
Even Vista is being reduced in such regard - way too many obtrusive 
security requestors to users. In other words, - don't bother end 
user with questions he knows sh*t about their meaning anyway:-)

Imagine if Google used the plugin for their ads - they would be able 
to store their whole database distributed amongst the computers of 
everyone on the internet. Would a security requestor be able to explain 
that to a newbie?

no

The advantage to cryptographic signing isn't just being able to track 
down an author, it also allows certificate revocation. With a free 
registry, revocation wouldn't matter - the bad guys would just register 
again.

askiing user e.g. what you discussed here - "do you want your previous 
cache to be deleted?" would result to "What is cache?" in 99% and 
users would press "Yes" .... or "no" .... :-)

This is why it would be best to use the browser cache for "let me 
store some graphics so I won't have to download them every time" 
situations. Other user settings are small in comparison, and can 
easily be stored in browser cookies or server side. Then, no security 
requestors necessary.

then we need to investigate ways of how to better utilise do-browser 
...

are we able to get to such browser settings in some unified way, 
so the same script works in all browsers?

You could write *-thru functions that used the browser to do the 
reading, with its cache.

read-thru, load-thru, etc., just like View uses for its sandbox.

hmm, that needs to be part of our user code. Not sure Carl will want 
to have two different versions of mezzanines - View, and plug-in 
one ...

There are two dlls here you know, the plugin calls the View dll. 
The browser-specific stuff is in the different plugin dlls.

my understanding is that it is the C code, wrapper for the browser. 
It would have to call View dll with some packed-in rebol code then 
to "preconfigure/patch" some mezzanine code ...