[Color Computer] Re: [Coco] C-Cubed

[Lothan]

I think this could work, but it won't be as simple as posting a file and
retrieving the binary (e.g. the HTTP POST/GET model). In order for make
files to work, the server should accept the make file as input followed by
requests of each source and library file required to build the project.

Perhaps a more intuitive approach would be to implement the client/server
logic in make itself. In this model, the HTTP POST/GET methodology still
works but make submits a request to compile each source file to an object
file followed by requests to link the files into a library or executable
module. There would be more traffic during the initial build, but subsequent
builds would (at least theoretically) be smaller incremental builds.

Putting the client/server logic in make rather than in the compiler
simplifies the compiler code and allows pieces of the compiler to be farmed
out to the server as required. The question then becomes which portions of
the compiler run on the server. If rlink were kept on the client, for
example, we wouldn't need to send potentially large libraries over an IP
connection and allows the server to become a relocatable object generator.
If rma were kept on the client, the server becomes a C to assembly
translator.

The advantage is that a configuration file could be used to specify which
compiler components are located on the server to allow the potential for
future "open source" compiler components that can be distributed. The
disadvantage, of course, is that all projects require a make file. How much
of a disadvantage this poses depends entirely on whether the developer
normally creates make files or uses the cc executive.

> -----Original Message-----
> From: coco-bounces at maltedmedia.com [mailto:coco-bounces at maltedmedia.com]
> On Behalf Of James Diffendaffer
> Sent: Thursday, July 21, 2005 8:49 AM
> To: ColorComputer at yahoogroups.com
> Subject: [Color Computer] Re: [Coco] C-Cubed
> 
> A remote hosted compiler is an interesting idea as an alternative to
> having to set up something like Cygwin (definately not for the
> beginner) or worry about copyright issues.
> 
> However... I'm not sure how well it will work.
> You must be able to 'make' a project with multiple source files,
> include files and your own libs.
> 
> 
> I see only a few ways this can work:
> 
> If you use standard include files and are compiling one source file.
> I guess it would work for a beginner but it would be of no use to a
> real developer.  No make capability.
> 
> Set up something like a unix account for them... not very practical.
> 
> Or they archive their entire environment with their own libs, includes
> and everything.  If anything were wrong with the directory structure
> in the archive it would fail but it's possible.
> 
> 
> I don't think anything for the CoCo will be big enough to cause a huge
> delay accessing such a system... but you would need to be online to
> use it.