Where is javaws.exe??

[Bent C Dalager]

It would work off the file extension. If you have a web site that has a jnlp
file that returns a text mime type the browser is still smart enough to know
to open WebStart. Then Webstart rejects it because it isn't the right MIME
type. I've done it before.

Interesting. Is this behaviour browser-specific or OS-specific in any
way? Is it specified in a standard somewhere?

Cheers
Bent D

 

[Dale King]

you have there about 100 lines of bubblegum that manages to persist
one value. Does that not seem ridiculous to you?

That bubble gum is more to do with implementing the preferences API than
using the persistence service. And note you can store more than one value.
You can store a whole hierarchy of values.

This is like some fancy car door lock that manages to befuddle the
owner into locking himself out. It often keeps the owner out, but
always permits in professional thieves.

And what you are suggesting is a tiny padlock that is totally inadequate for
any but the most trivial applications.

That bailing wire scheme is highly inconvenient and further it offers
no more security that had the unsigned app been told it could write
only X bytes of files in the current directory.

Such a limitation on access cannot be done with the current I/O scheme.
There is no means for allowing only access to a given directory and limiting
how much you can write. It would require some new API instead of the normal
File, FileInputStream, and FileOutputStream. Such an API would not be
drastically different than the Persistence service.

Your scheme would lose the ability to share data between different apps from
the same host. That is a design feature of the PersistenceService. And don't
forget the ability to "tag" data as CACHED, DIRTY, or TEMPORARY.

 

[Dale King]

and what pray tell is that number.? This is the Mickey Mouse heart I
bitched about.

That is a number requesting the given amount of space. This allows the JNLP
client to reject the allocation as too large before the data is written
rather than waiting to throw an exception in the middle of writing. This way
of doing it is more robust.

 

[Dale King]

It is not a security hole if you put a limit on its size and give him
his own directory to play in and don't let him out.

Exactly, that is exactly what the PersistenceService does.

The advantage is then JWS apps can function without modification as
ordinary apps, just by installing them with ZeroG or the like.

Once again you could not put such limitations on him with the normal I/O
calls.

A malicious app can likely do more damage with the API presented by
filling it with junk. At least with a disk scheme, you could have a
clear quota controlled by the policy file.

Then please explain how that is possible. There are security premisssions
that can restrict what portions of the file system you have acess to, but
there are no security permissions dealing with how much data you can write
to a file. And we are worried with not just quotas on an individual file,
but quotas on the entire space allocated to the app.

 

[Dale King]

Interesting. Is this behaviour browser-specific or OS-specific in any
way? Is it specified in a standard somewhere?

I don't know. It does it with IE under Windoze. Even if it is not universal
relaxing the restriction will still help. Remember that you can now enter
URL's directly into WebStart so even if the browser association did not work
that is not a big problem.

 

[Bent C Dalager]

I don't know. It does it with IE under Windoze. Even if it is not universal
relaxing the restriction will still help. Remember that you can now enter
URL's directly into WebStart so even if the browser association did not work
that is not a big problem.

Assume for the sake of argument that it works on Windows but not on
Mac. What we'd end up with then are JWS apps that would work on
Windows machines but not on Macs. This seems to me to be counter to
the idea that they should work everywhere. If this is the case, then I
think it is better if JWS is as restrictive as it is today since it
forces developers to adopt a cross-platform solution to the deployment
problem. Even if it _does_ require you to send enforcers over to your
ISP in order to get them to fix their mime support files

Cheers
Bent D

 

[Bent C Dalager]

That is a number requesting the given amount of space. This allows the JNLP
client to reject the allocation as too large before the data is written
rather than waiting to throw an exception in the middle of writing. This way
of doing it is more robust.

I currently prefer to ask for just the amount I know I'll need and
then resizing it later if it should become necessary.

Cheers
Bent D

 

[Dale King]

Assume for the sake of argument that it works on Windows but not on
Mac. What we'd end up with then are JWS apps that would work on
Windows machines but not on Macs.

No, you would end up with JWS apps that would not be started automatically
by clicking on a link in a browser on some machines. Which by the way is
exactly what we have now considering that it doesn't currently set up the
association in browsers like opera or on Linux. The app would still work you
just have to cut and past the link over to the WebStart app.

This seems to me to be counter to
the idea that they should work everywhere. If this is the case, then I
think it is better if JWS is as restrictive as it is today since it
forces developers to adopt a cross-platform solution to the deployment
problem. Even if it _does_ require you to send enforcers over to your
ISP in order to get them to fix their mime support files

But means that some people (including me) cannot publish WebStart apps at
all because our ISP does not support the mime type.

 

[Bent C Dalager]

No, you would end up with JWS apps that would not be started automatically
by clicking on a link in a browser on some machines. Which by the way is
exactly what we have now considering that it doesn't currently set up the
association in browsers like opera or on Linux.

Which problem is this? I am using Opera on Windows and haven't had any
trouble getting click-to-run to work with JWS. I haven't tried from my
Linux box yet. I probably should.

But means that some people (including me) cannot publish WebStart apps at
all because our ISP does not support the mime type.

I can see how this is a problem, but I fear your proposed cure will
end up doing more harm than good.

Cheers
Bent D

 

[Dale King]

Which problem is this? I am using Opera on Windows and haven't had any
trouble getting click-to-run to work with JWS. I haven't tried from my
Linux box yet. I probably should.

Sure, you can get it to work, but it must be set up manually. You can do the
same manual set-up to associate jnlp extensions with WebStart.

I can see how this is a problem, but I fear your proposed cure will
end up doing more harm than good.

And I know that there is already harm. I have a WebStart app or two that
with a little work I could publish, but I lack a good place to do so. I
don't want to pay for web hosting for free apps when I already pay for an
ISP as part of my cable modem. So for lack of a good place to publish them
they sit in limbo. All because Sun is overly restrictive.

 

[Bent C Dalager]

Sure, you can get it to work, but it must be set up manually. You can do the
same manual set-up to associate jnlp extensions with WebStart.

This is strange. I can't remember having done anything in particular
to get it to work, other than install the VM. I would have thought
that the installation program registers the file and mime type with
the Windows "file types" thingie and that Opera just uses this?

Cheers
Bent D

 

[Roedy Green]

I would have thought
that the installation program registers the file and mime type with
the Windows "file types" thingie and that Opera just uses this?

I think Opera now sets up the association for you automatically, or if
not at least inherits the association from an earlier9 version set up
manually. See http://mindprod.com/jgloss/opera.html for how to set up
associations manually.