[REBOL] Re: ANN: rugby 4.3
From: petr:krenzelok:trz:cz at: 21-Nov-2001 15:54
wooohooo, well, let's dig into it ...
Maarten Koopmans wrote:
> Hi all,
>
> I just released Rugby 4.3 , the best version ever. All upgrade!
>
> Download it at http://www.rebolforces.com/~erebol/
>
> Rugby is a rebol request broker that allows you to build distributed
> computing applications in a few lines of code! A great example is Graham's
> GoRim.
Yes, it is. However, let's prove we are right. I don't call GoRim a success yet,
nor do I call RIM a success. I think that polling aproach for instant chat is the
wrong way to go.
I don't say Rugby is bad architecture, as it is not and it has its future with
Rebol! However - does its transport mechanism allow us to build something more
real-time, in quality level of IRC? Does IRC use polling? Why should XYZ users
contact/poll server each N secs, if there is no message available?
Now - would it be possible to build another scenario using Rugby? Frank showed us
with his IRC client, that communication using Rebol CAN be really instant.
We currently can -
- call remote function (via imported stubs,or directly remotely)
- call remote function /deffered - is connection closed here?
- call remote function /oneway - is connection closed here?
- call remote function /wait-for-result - does this block whole user interface,
till we don't receive answer?
What about following:
- chat client running rugby client and server.
- chat client connects to server
- server registers client's listening ports, and connects to it, keeps the
connection *
- once message arrives to server, it distributes it to all client channels ...
*But probably, we've got problem with firewalls here ...
It's not critique, just reality - I showed GoRim to several friends, and typica
answer was - why is it so slow in response?
What is your opinion on the subject?
>
> What's new:
>
> - Transparent error propagation. An error on the server is now serialized and
> instantiated on the client as the same error! This makes it easy to do
> distributed debugging.
>
Be carefull with error ;-) I pasted Graham one error message, which showed whole
path in my system on my server. I call it a security risk :-)
> - Http server side handler is improved in the way it clears the buffers at
> the end of a request. This may (or may not) fix the runaway problem after a
> long period of time where servers start eating CPU time on some platforms.
>
great!
> - Functions now support refinements! YES!
>
> - Totally renewed stub generation engine for refinement support. As a side
> effect you can now import multiple services in the global namespace.
>
> - Some minor performance improvements compared to 4.3e resulting in a 7-10%
> increase overall.
>
> I will add extra documentation in the next few weeks, although the current
> documentation should be good enough for most of you. Please don't hesitate to
> ask me questions by email as specfied on the web site.
>
Everything great!
-pekr-
