Re: [gomd-devel] <LIBRARY>: work started

[Matthias Rechenburg]

Hi JP,

On Sonntag 08 Juni 2003 01:12, Gian Paolo Ghilardi wrote:
> Hi all.
>
> As discussed on #gomd with Mirko (busybee) and Marco (CaloRE), I've just
> started working on libgomd.
>
> Sorry Matt: you're the maintainer of libgomd. So you have the power to
> annihilate my patches... :)

no sorry ;) go ahead.
I am sorry for less time.

>
> The idea is the following (to be implemented):
> 1)A new class ("netCode", my fantasy lacks at the moment) includes a few
> functions to interact with gomd

yep

> 2)openConn(IP, port): tries establishing a conn with a gomd daemon

yep, IP + port read from a config-file.
I think we should provide a standard config file for the libgomd (and libgomd
client apps). The mechanism should be :
- try to read config from file (.gomdrc, /etc/gomd.conf)
- if not found try localhost : default port

We may need another config entry in /etc/gomd.conf -> master-gomd-ip
Then we can use this config-file for the libgomd-apps too and we may
need this entry anyways for the gomd2gomd methods.

And, sure, the gomd does not need to run on the same system than
the later libgomd client app, the should just use the same config-file,
or at least the same config-file-syntax.
So later the libgomd developer will just install its application and
a ~/.gomdrc or a /etc/gomd.conf and that is it. If none of these
config files where found the libgomd application will try to connect to
localhost.

Any comments or better ideas ?
;)

> 3)requestInfo(infoToRequestStr): if the conn was established, requests
> something to a gomd daemon and wiats for the standard one-line-only answer.
> If something goes wrong (i.e.: the socket timeout error occurred) returns
> an error message. Otherwise it return the answer from the daemon.

yep

> 4)closeConn(): closes the conn (if established).

yep again.

> 5)Each of these C++ functions calls a specific C function: C++ functions
> don't do anything. They call and return values from C-functions placed
> outside of class-definition. So C++ functions are only wrappers to C
> functions.
> This approach allow us to provide a structured-style for C clients and an
> OOP-style for C++ clients (without writing two distinct libraries).
>

very nice. I agree with all of you suggestions.

> Please notice the point #3.
> The idea is to make clients' life easier.

exaclty.

>
> Comments?
>
> <rejected>
>
>
>
> _______________________________________________
> gomd-devel mailing list
> address@hidden
> http://mail.nongnu.org/mailman/listinfo/gomd-devel

all the best,

Matt
-- 
E-mail  :  address@hidden
www     : http://www.openmosixview.com
an openMosix-cluster management GUI

" I'm gonna drink till I Re-boot..."