Hi,

Am Dienstag, den 06.05.2008, 06:10 +0200 schrieb David Nuescheler:
> > > I am fine with this, too. If there is no objection I will change that
> > > value to trigger the IE-friendly (which it is ;-) ) mode to "browser".
> >  I don't see any reason to include it as a mode.  As a workaround, sure,
> >  but that should be very specific to the User-Agent field and not require
> >  special programming by application devs.

>From an ease of use POV, I 100% agree with Roy, but ...

> Actually, this is not about a specific browser, this is the only way
> how any browser can deal with a file multipart upload so a dhtml
> app can access its result.

The problem lies in some browser's "ability" to allow for kind of a
user-friendly error message. Instead of relying on the the server to
provide user-friendly messages, the browser display something they think
is more user friendly. (Granted most error responses on the real-world
net are not user-friendly, mostly filled with stack traces or DB error
messages ....)

The concrete case is IE which allows setting a "Show friendly HTTP error
messages" message flag (in the Advanced Tab of the Internet Options),
which hides the actual server response.

There is no way a server may find out about this situation. Hence the
workaround is for the client to set a flag to simulate a successfull
request and send back the actual status in the resulting HTML.

As this is really a workaround, which is used in one specific use case,
it must be forced by setting a flag. Otherwise - and by default -, the
server will act standards-complying by sending back the correct status
code together with the full status response.

Regards
Felix

> 
> Since there is no client side trick (for security reasons) to do a
> file upload other than a straight form POST (no xhr or anything)
> and since there is no response code other than 200 and no
> other way than the DOM of a text/html document that allows you
> to transport the result of the POST (like let's say the "location"
> after the 201) to a client-side app this is not a browser specific
> workaround. I wish there was something else but browsers
> are very consistent here. Frankly, I am not even sure what
> browsers could do about that in the future.
> 
> regards,
> david

Reply via email to