Index: [thread] [date] [subject] [author]
  From: Daemonprince <dp@khazad-dum.freeserve.co.uk>
  To  : ggi-develop@eskimo.com
  Date: Tue, 23 Feb 1999 23:33:48 +0000

Re: libGGI3d -- ready to code?]

On Tue, Feb 23, 1999 at 08:23:45AM -0500, Filip Spacek wrote:
> The way I see it is impossible to assign the same ids to two modules with 
> the same interface without some sort of parser. For example if you have
> two triangle drawers, you will get two _different_ ids. We will have to
> have some sort of parser to recognise during compilation that the two
> interfaces are indeed identical, or during pipeline construction to figure
> out that those two ids represent the same type of interface. This parser
> would do pretty much the same thing as a parser for non-COM proposal so
> basically I don't see much advantages in using COM (except for runtime
> shuffling in pipeline)
> 
Nah, thats the idea behind COM and CORBA and such like. The same interface
has the same Id no matter where it is. When you need to find something to
service you, you just ask what has Interface X. There is no need to identify
which module is which because they are seperate files.

I think a COM like interface would work like an extended ELF interface from
what little I know about ELF. Request Interface Id X get served a possible
list of modules. Select fastest one. This goes on down the pipe and this
way we are profiled and pipe is built. Smart modules which do things
like threading can be stuck into this pipe to send stuff down alternate pipes.

The thing I can see definately being needed is an atomic write to both the
Z and pixel buffers. Though this may be acheived by having all branches
rejoin at final level.

Graeme


----- End forwarded message -----

Index: [thread] [date] [subject] [author]