Re: [OT, Q] Fast RPC library?

[Richard Frith-Macdonald]

On 2005-10-23 23:54:41 +0100 Helge Hess <helge.hess@opengroupware.org> wrote:

> On 23. Okt 2005, at 12:14 Uhr, Richard Frith-Macdonald wrote:
> > One option, if you trust the clients to generate the valid xml, is  to 
> > write your own extremely simplified xml parser.
>
> Well, but then, what would be the point in using XML-RPC? The only  real 
> gain 
> you have is the implicit advantages of using a standard  which implies that 
> you have to deal with all the stuff attached.

Clients could use standard libraries ... so it would be easy people to 
develop the client apps, but the server could run faster.  I've seen one 
very large company take this approach in order to support high volumes of 
RPCs sent by their partners.

> What I can imagine is that pure HTTP with some custom XML format  might be 
> an 
> option. With a proper HTTP/1.1 library which is  supporting keep-alive and 
> pipelining, one might get the requested  amount of RPCs/s while still 
> having 
> a scalable and standards  compliant base.

I use the WebServer class for that.  I'm sure there are plenty of other 
options too.

> Actually I wonder _why_ you want to replace your existing system if  that 
> works?

He said his existing system is 'platform dependent binary protocol'  so I 
guess to move away from platform dependency?

> > The GNUstep SQLClient library (in the dev-libs directory) contains  the 
> > WebServer class
>
> Why the heck do you put a webserver class into a SQL library?! ;-)

Historical mistake ... started off as a tool to provide a web-based database 
access test utility ... needs to be separated out into it's own .library 
when I have some time with nothing better to do.
.