Dear all,
from the Client point of view I completely agree with Thomas about doing things as simply as possible, as initially planned. This can encourage developers to use the Plastic technology and get on board as soon as possible, giving the possibility to work on what really matters: messages. Orphaned files can be deleted after a proper check of their status. For example, when VOSpec tries to connect to the hub and fails I check if there is a dot file. If there is no dot file I start an internal one, if there is a dot file I just throw an Exception. Of course here I could be a little bit more intelligent and delete it.. I also agree with John that transferring complexity to the server side is the best approach. We are currently few applications using Plastic and we didn't find serious problems for the lack of elegance.
At the end simplicity was one of the first Plastic goals and I definitely propose to stay on this line as much as possible.
Best Regards,
Isa Barbarisi
Thomas Boch wrote:
> Hi all,
>
> As for the bootstrap process, I'm still not convinced that we need a
> socket-based approach. Sure, it is more elegant than the file-based one,
> but to me, this elegance does not counterbalance the additional
> complexity at the client side.
>
> Are orphaned files really a problem ? If an application reads such a
> file, tries to connect to the hub and fails, it sounds sensible to
> consider this hub as dead, to delete the dot file, and start up a new
> hub.
>
> I think that one of the reasons of the rapid take-up of PLASTIC is its
> simplicity. I have nothing against designing a complex architecture, as
> long as we gain benefits from this complexity.
>
> Cheers,
>
> Thomas
>