Re: [Plastic-devs] plastic file ref

[R H. <ric...@co...>]

On Tue, 4 Jul 2006, Mark Taylor wrote:

> On Tue, 4 Jul 2006, R Holbrey wrote:
>
>>> To prevent this kind of confusion though, maybe there should be something
>>> explicit in the message definition for loadURL which says that the URL is
>>> only guaranteed to remain usable between when the message is sent and
>>> when the receiver provides the return value (or possibly some other
>>> policy we agree on).
>>>
>>> Mark
>>
>> Wouldn't work for me either now I think about it - I'm relying on a
>> persistant copy of the file somewhere. I guess it's up to me to make a
>> copy, but topcat etc could get very bored waiting for replies to come
>> back. If you were sending to a specific receiver, might it be possible
>> to transfer ownership?
>
> Not inconceivable, but it would start to make that message very messy,
> especially since it may not be a temporary file, or a file at all
> (may well be an http: type URL etc).  To follow that route probably
> the best thing would be to have the hub providing some service to do
> the hard work - however, like John, I'm not very enthusiastic for the
> hub to start broadening its portfolio like that.
>
> I appreciate there is a problem, though it's probably not possible to
> solve it with complete reliability (staged SIAP tables etc).
> If you really do need a persistent version of the table then probably
> making a copy is the right thing to do, though it's not very elegant
> since a broadcast might result in many listening applications making
> (possibly unnecessary, if it's not a temporary file URL after all)
> copies of the same data.  However as far as response time for the
> message goes I'd say it is reasonable for Eirik to take time to do
> the copy before replying.  An alternative hack, though it's getting
> a bit OS-specific, is to get some sort of handle on the file which
> will allow you to continue to access it even if it's deleted
> (in Un*x you could make a hard link, or just keep a copy of the
> open file handle).
>
> Mark

I still see an admin role for the hub here (not popular, I know). It now 
see it would mean extra work and complexity, but hey, why write the code 
there when it can be done again and again in Java, c++, python..   ;-)

Also, I suppose badly behaved apps might not choose to respond and then 
the hub has to start making decisions... but then I thought I observed 
that the hub unregistered eirik, if it returned "false" ?

Richard